<?xml version="1.0" encoding="UTF-8"?><rss xmlns:dc="http://purl.org/dc/elements/1.1/" xmlns:content="http://purl.org/rss/1.0/modules/content/" xmlns:atom="http://www.w3.org/2005/Atom" version="2.0" xmlns:cc="http://cyber.law.harvard.edu/rss/creativeCommonsRssModule.html">
    <channel>
        <title><![CDATA[Stories by Junta Sekimori on Medium]]></title>
        <description><![CDATA[Stories by Junta Sekimori on Medium]]></description>
        <link>https://medium.com/@juntaah?source=rss-8cd5173a44bc------2</link>
        <image>
            <url>https://cdn-images-1.medium.com/fit/c/150/150/1*c4OXU8_mPC_FKpy88u0oKw.jpeg</url>
            <title>Stories by Junta Sekimori on Medium</title>
            <link>https://medium.com/@juntaah?source=rss-8cd5173a44bc------2</link>
        </image>
        <generator>Medium</generator>
        <lastBuildDate>Fri, 20 Nov 2026 16:21:21 GMT</lastBuildDate>
        <atom:link href="https://medium.com/@juntaah/feed" rel="self" type="application/rss+xml"/>
        <webMaster><![CDATA[yourfriends@medium.com]]></webMaster>
        <atom:link href="http://medium.superfeedr.com/" rel="hub"/>
        <item>
            <title><![CDATA[An interactive explanation of Bayes’ billiard table experiment]]></title>
            <link>https://medium.com/@juntaah/an-interactive-explanation-of-bayes-billiard-table-experiment-78fb34eb7bc1?source=rss-8cd5173a44bc------2</link>
            <guid isPermaLink="false">https://medium.com/p/78fb34eb7bc1</guid>
            <category><![CDATA[data-science]]></category>
            <category><![CDATA[statistics]]></category>
            <category><![CDATA[vibe-coding]]></category>
            <category><![CDATA[analytics]]></category>
            <category><![CDATA[bayesian-statistics]]></category>
            <dc:creator><![CDATA[Junta Sekimori]]></dc:creator>
            <pubDate>Mon, 13 Apr 2026 06:02:53 GMT</pubDate>
            <atom:updated>2026-04-13T06:04:58.744Z</atom:updated>
            <content:encoded><![CDATA[<h4>And how anyone can vibe code web apps for more impactful data science communication</h4><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/0*BsATTCGxQpy6rNQt.png" /><figcaption><a href="https://juntaah.github.io/bayes_billiards/">Link to the interactive Bayes simulator and betting game</a></figcaption></figure><p>AI-assisted coding gives us some exciting opportunities for more impactful data science communication, and in this article I want to share a web app I created to help understand Bayes’ billiard table experiment.</p><p>Being able to go from vision to a functional app in such a short space of time has opened my eyes to how accessible front-end coding has become and how ubiquitous interactive tools will surely become in education. Any educator with an idea can now build a simple app at no cost and with just basic technical knowledge.</p><h4><strong>What is Bayes’ billiard table experiment?</strong></h4><p>Inference is the art of making good judgments from the information available to us. And as new information comes in, we need a good way to calibrate our judgments. This is the gist of Bayesian thinking and what the famous billiard table experiment tries to illustrate.</p><p>It goes something like this: the cue ball is placed somewhere on the table unknown to you. You need to guess its position. To help you with your guess, you get to roll balls (without looking at the table) randomly onto the table and be told whether it settled to the left of the cue ball or to the right.</p><p>The idea is that with every new roll of a ball, you get more information to help you with your guess. If in the first 5 rolls you had 4 land to the left and 1 to the right, you would be reasonable to guess that the cue ball is on the right side of the table. If after 50 rolls, you had 40 on the left and 10 on the right, you would feel confident that the cue ball was around 4/5ths across towards the right side of the table.</p><p>What we see is that, with each roll of the ball, we are able to update our prior knowledge (which begins with a totally uninformed state — or ‘uniform priors’) with the new information and calibrate our guess with the new data.</p><p>Matt Parker and Hannah Fry have <a href="https://www.youtube.com/watch?v=7GgLSnQ48os">a fun video on it</a> if you can bear watching him repeatedly miss the table with his ball throws. Equally, you can pick up practically any book that delves into a fundamentals of Bayes and this is likely to be explained (try Chapter 11 of <a href="https://medium.com/@juntaah/review-the-art-of-statistics-david-spiegelhalter-ed366f4102ea"><em>The Art of Statistics</em></a><em> </em>by David Spiegelhalter, for example).</p><h4>The app</h4><p>The app is available <a href="https://juntaah.github.io/bayes_billiards/index.html">here</a> and comes in two parts, which can be toggled from the top of the page:</p><ol><li><a href="https://juntaah.github.io/bayes_billiards/index.html"><strong>The simulator</strong></a> which shows for each ball roll (though in this version, we are throwing them), how the probability distribution (and most likely position of the secret ball) changes. The more the throws, the narrower the distribution and the greater precision of our estimate.</li><li><a href="https://juntaah.github.io/bayes_billiards/bayes_game.html"><strong>A betting game</strong></a> based on the same concept which puts you in the hot seat of guessing the position of the cue ball. The quicker the guess, the greater the opportunity for payout!</li></ol><p>Hopefully, the interface is intuitive to you and you get the same satisfaction I get from watching the distributions change.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/0*6zWSyGT-hlJWOqQu.png" /></figure><p>Use the Simulator tool to see the relationship of where the cue ball (I have coded this as a ‘red ball’ for ease of display) actually is vs where the evidence from our ball tosses suggests the ball after any given ball toss. You will notice that the red ball should be within the 95% ‘credible interval’ about 95% of the time, and that it is rare for the peak of the distribution (in frequentist terms the ‘maximum likelihood estimate’ which is used to make point estimates) to line up with the actual position of the ball.</p><p>The betting game is designed to bring us into the mindset of a real-life decision making and get a more intuitive feel for how we would actually update our beliefs from new evidence. While this toy example is trivial, the app helps to show us the framework by which could tackle more pressing problems like the position of missing ships, or other situations in which we have to make decisions under scant data.</p><h4>How I built it</h4><p>I built the app with Gemini 3 (Fast). Gemini Canvas provides an interface more dedicated to code generation but I used the simple chat interface for no better reason than it just worked and I got into the flow of it.</p><p>The first prompt was <em>Build me an interactive app where I can play the guessing game where Bayes threw balls on a billiards table </em>which got most of the core functionality in one go. Gemini wrote the code in HTML in a React framework, which I could simply save as an html file and open in a browser to immediately see.</p><p>What it generated was already exciting but there were specific UI elements that felt awkward to me and features I felt would improve it such as the ability to show and hide where the thrown balls landed. These were added through a number of follow-up prompts across the next hour or so.</p><p>Hungry to test my own skills of inference after getting to see the posterior distributions move around, I then skipped dinner to build out the betting game in a second tab. This too it got most of the way after the first prompt which looked like this:</p><p><em>Give me a second version of this page where the user now gets to play a betting game. They start off with $100. They get 20 games and before each game, the user wagers any amount of money they have remaining. And after each throw they can choose to make a bet on where the red ball is by moving the slider on the axis. If they are within 0.1 distance of the red ball, they win. The payout amount is proportional to the precision of the posterior distribution. If the precision dips below 0.08, the wager is lost and the game ends. Correct guesses when the distribution has a precision greater than 0.08 get a payout proportional to the precision. The lower the precision, the higher the payout. Keep the layout scheme similar to the original page. The distribution graph should appear only at the end of each game.</em></p><p>This one took a little longer to customise to my liking, and the final version uses different parameters and rules to my original ideas. I would say the first prompt got me 90% of the way in the first 5 minutes. Working on the next 10% to make it just right took me a couple of hours across another couple of dozen prompts.</p><p><strong>Deployment of the app was easy through Github pages</strong>. There is nothing to pay. Just create a new repository with these web pages, go to the ‘Pages’ section from the settings and deploy it. It took just a few minutes all in all.</p><p>I left the two pages as flat files which I’m sure seasoned developers would wince at but I should think it wouldn’t be too much of a leap to convert this to a tidier repository structure through Antigravity, Claude Code or the like.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/0*_q0VsxFexOLX3YIE.png" /></figure><h4>Thoughts on the future</h4><p>I have enjoyed writing about <a href="https://towardsdatascience.com/how-to-intuit-the-prosecutors-fallacy-and-run-better-hypothesis-tests-2d8561bfc5a6/">Bayesian</a> <a href="https://towardsdatascience.com/how-to-tackle-the-weekend-quiz-like-a-bayesian-b5e035ba3746/">topics</a> in the past, but experiencing how quick it was to get this up and running <strong>makes me excited to explore interactive apps as a communication medium open to me as an alternative to writing</strong> or — in more of a work context — putting together slides. I can see myself being able to get my points across quicker this way in certain situations, such as in presentations of technical concepts to business stakeholders.</p><p>The pace at which I was able to have a vision and put something like this together so quickly was eye-opening, and I think the opportunity for educators to incorporate interactive tools like this in their toolkit is game changing.</p><img src="https://medium.com/_/stat?event=post.clientViewed&referrerSource=full_rss&postId=78fb34eb7bc1" width="1" height="1" alt="">]]></content:encoded>
        </item>
        <item>
            <title><![CDATA[Review: Making of a Manager, Julie Zhuo]]></title>
            <link>https://medium.com/@juntaah/review-making-of-a-manager-julie-zhuo-5bde85784372?source=rss-8cd5173a44bc------2</link>
            <guid isPermaLink="false">https://medium.com/p/5bde85784372</guid>
            <category><![CDATA[analytics]]></category>
            <category><![CDATA[books]]></category>
            <category><![CDATA[management]]></category>
            <category><![CDATA[book-review]]></category>
            <dc:creator><![CDATA[Junta Sekimori]]></dc:creator>
            <pubDate>Tue, 17 Mar 2026 11:41:42 GMT</pubDate>
            <atom:updated>2026-03-18T01:14:29.824Z</atom:updated>
            <content:encoded><![CDATA[<h4>A vivid picture of what it’s like to work as a manager of people in analytics and an easy recomendation for new managers.</h4><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*_u5Iu9WOe5UApx8UBpqMTQ.jpeg" /></figure><p>The greatest practical tool I have taken away from <em>Making of a Manager</em> is the PPP mnemonic, which I have found my mind wandering back to across my years managing analytics teams.</p><p>According to the author, who captures her experiences learning to manage teams at Facebook in this memoir-like book, the role of a manager fits into three neat buckets:</p><ol><li>Purpose</li><li>People</li><li>Process</li></ol><p>If ever I am at a loss as to how to use my time, or need to bring order to my Tuesday afternoon amid a jumble of ideas and anxieties as a manager (a state <a href="https://fs.blog/matthew-crawford-shop-class/">Matthew Crawford</a> memorably called a <em>bewildering psychic landscape</em>), referring back to this PPP often gives me the clarity I need to regain sight of the things that matter. If in doubt, think which P could most benefit from attention right now.</p><p><em>Purpose </em>is all about what your team should direct their focus towards and why they should want to do a good job at it: “<em>The first big part of your job as a manager is to ensure that your team knows what success looks like and cares about achieving it.” </em>(p.22) If the team’s work is not having clear impact on the organisation, what could they be doing differently and what can you do to rally your team to this vision?</p><p><em>People </em>is about whether your team has the right people for the purpose and whether there is anything you can do to set them up better for success. Are there any capabilities missing in the team? Are these best introduced through hiring or coaching? What does a great ‘culture’ look like in your team and what can you do to improve this?</p><p>Finally, <em>process </em>is about <em>how</em> your team carries out the work. What are the factors slowing down or reducing the effectiveness of your team? What rules or guidelines can you put in place to mitigate these?</p><blockquote><em>Often people have an allergic reaction to the word process… In a processless world, I imagined myself free to do whatever was needed to make things happen quickly, with no red tape, no barriers, no overhead… In a team setting, it’s impossible for a group of people to coordinate what needs to get done without spending time on it. The larger the team, the more time is needed… For managers, important processes to master include running effective meetings, future proofing against past mistakes, planning for tomorrow, and nurturing a healthy culture.” (p.24)</em></blockquote><p>I love the simplicity of PPP and, while not every minute of your day will be spent on a P (what about putting out those fires or reviewing a deliverable?), it’s a useful baseline to return to when building or managing teams. Over the period of a quarter, if I am not spending more time on PPP than other concerns, then I need to make adjustments. It’s the first thing that comes to mind when I am asked for advice from new managers.</p><p>Another memorable heuristic the author conjures is to optimise your time towards being a <em>multiplier </em>for the team as opposed to having an additive effect.</p><blockquote><em>“Your role as a manager is not to do the work yourself, even if you are the best at it, because that will only take you so far. Your role is to improve the purpose, people and process of your team to get as high a multiplier effect on your collective outcome as you can.” (p.26)</em></blockquote><h3>Every manager feels like an imposter sometimes</h3><p>The book is written in a conversational and unacademic style, very different to, say, HBR articles that cite data to back up theory. This is friendly guidance rooted in the lived experience from a relatable introvert (a designer) who learned to lead by necessity. After the stomach knots and tears of her own journey, this book serves as the hard-won culmination of the lessons she can now share.</p><p>But what makes her the authority on the subject that makes the advice worth applying? Her main claim is that she was a VP at Facebook, having worked her way up from a grad. Re-reading the book now, the undertone of working at Facebook as the self-evident pinnacle of professional success is a bit grating (probably because I am a grumpy social media curmudgeon), but the writing is grounded and there is a lot here to relate to.</p><p>The book expands on <em>people </em>a lot more than it does on the other two P’s. If you are stuck as to how to drive <em>purpose </em>in particular, you will find little here to reveal the way. What it does very well, rather, is to recognise the difficulties new managers face in adapting to the new way of thinking generally and making this feel like part of the journey. It’s ok to feel like you’re out of your depth. It’s ok if this job feels too big for any one person to handle. It’s ok to not get it right all the time. It’s all part of the learning.</p><p>My favourite chapter is on <em>Managing Yourself</em>. It says something about the management experience and the author’s <em>purpose </em>that this is the biggest chapter, bigger than <em>What is Management </em>and <em>Making Things Happen. </em>It can feel like you are always on show as a leader and you need to find ways to manage yourself to push through the doubt.</p><blockquote>“<em>Over the years, I have learned a secret that bears repeating: Every manager feels like an imposter sometimes. Every manager was once new, stumbling through interviews and 1:1s and awkward conversations. It’s so common that instead of pretending like we are all ducks gliding effortlessly on the surface of the water, we should own up to the furious paddling that is happening beneath</em>.” (p.109)</blockquote><p>The chapter focuses on the idea that management is a very personal journey and that we need a good understanding of ourselves — our strengths, values, blind spots — as a foundation to being successful long term. It calls on really getting to grips with who we are at our best and at our worst and building the confidence to push through when the going is tough.</p><blockquote>“<em>No matter the twists and turns of your managerial road, on some days you will feel the imposter syndrome so strongly that you might as well be stuck at the bottom of a dark, deep pit. Every manager I know is well acquainted with this place, where the scathing commentary of your inner critic echoes off the sheer walls, turning whispers into screams. In the Pit you feel so very alone</em>.” (p.122)</blockquote><p>The advice in the chapter is very compelling, due in part to the vulnerability she brings. You get a really raw sense of the times she spent in ‘the Pit’ and how pragmatically she worked on her own mindset and sought advice. It’s in this chapter also that she claims “<em>management isn’t an innate skill</em>” and that “<em>you must actually go through the real thing in order to gain the experience you need</em>”, which could almost be the first line of the blurb on the back page. In learning to coach yourself in this way, I think you’re also developing the wisdom to coach others too.</p><p>I love recommending this book to aspiring managers. It isn’t science but I can’t think of any other resource that paints a more vivid picture of what it’s actually like to work as a manager.</p><h3>Stray notes</h3><ul><li><em>The most important thing to remember about hiring is this: hiring is not a problem to be solved but an opportunity to build the future of your organisation</em>. (p.164)</li><li><em>Beyond people, you and your report should be aligned on why you’re doing what you’re doing and what success looks like. As Antoine de Saint Exupéry has been attributed as saying, “If you want to build a ship, don’t drum up the men to gather wood, divide the work, and give orders. Instead, teach them to yearn for the vast and endless sea.”</em> (p.282)</li><li><em>Rule of thumb of delegation goes like this: spend your time in the intersection of what’s most important to the organisation and 2) what you’re uniquely able to do better than anyone else. </em>(p.232)</li><li>It’s great to see Julie Zhuo continuing to write constantly. Some great essays on work in the era of AI: <em>Today’s AI does a far better job on many of these above dimensions than I do. It writes specs more clearly. It supplants hypotheses with research. It drafts 10x faster than me. If I find myself wary about harnessing its strengths, it’s only because of my pride. And yet. When I write to think, AI can only play a supporting role. The entire process of putting what I know in front of me IS what makes my brain sharper. There is no shortcutting this process. And when I write to feel, what would be the point of using AI for that? It can’t feel my feelings for me. The same can be said for coding. Maybe you hand-code on a Sunday afternoon and it’s your relaxation time. Great, no reason to give that up. (</em>from <a href="https://joulee.medium.com/the-thing-you-are-expert-at-will-be-your-career-downfall-005400ce6888">The Thing You Are Expert at Will Be Your Career Downfall</a>)</li></ul><img src="https://medium.com/_/stat?event=post.clientViewed&referrerSource=full_rss&postId=5bde85784372" width="1" height="1" alt="">]]></content:encoded>
        </item>
        <item>
            <title><![CDATA[Review: Cracked It!]]></title>
            <link>https://medium.com/@juntaah/review-cracked-it-56060aacfe39?source=rss-8cd5173a44bc------2</link>
            <guid isPermaLink="false">https://medium.com/p/56060aacfe39</guid>
            <category><![CDATA[books]]></category>
            <category><![CDATA[analytics]]></category>
            <category><![CDATA[business-strategy]]></category>
            <category><![CDATA[problem-solving]]></category>
            <category><![CDATA[consulting]]></category>
            <dc:creator><![CDATA[Junta Sekimori]]></dc:creator>
            <pubDate>Sat, 20 Dec 2025 12:42:33 GMT</pubDate>
            <atom:updated>2026-03-17T11:49:15.252Z</atom:updated>
            <content:encoded><![CDATA[<h4>Beyond SQL: why technical excellence isn’t enough to drive real world change</h4><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*Dco1hsaKksuqGGBAO7mZ-A.jpeg" /></figure><p>Management consultants.</p><p>I was cynical of them for at least the first decade of my career in analytics. That was generally the take among those who ushered me into my profession in the 2010s, and an early encounter in London with a Big 4 consultant who bragged about overwhelming their clients with 100 slides only seemed to confirm the image: <em>bullshit merchants</em>.</p><p>Looking in from the outside, the skillset of the management consultant was vague to me. Compared to our hard-earned technical specialism to manipulate data, what is it that these consultants provide that others can’t? I had come across examples of ‘frameworks’ they teach in business school but they seemed childishly simple compared to the mountainous undertaking of learning to code.</p><p>Coming around to the value of these frameworks and to the value a good consultant can bring has been a defining theme of the second decade of my career. Transitioning into analytics management, I was forced to reckon with the stress of <em>not </em>using frameworks and now, several years into this journey, I find myself hungry for useful frameworks in the same way that I lapped up coding tutorials earlier in my career.</p><p>I recognise now that the skill of a great consultant is the skill of solving problems <em>in the real world</em>, amidst the politics, uncertainty and time constraints of working in a living organisation. In the real world, there are rarely certain answers like there are in the classroom, and for both consultants and the analytics profession alike, excellence comes down to how effectively we can influence decisions within the context of all the considerable adversities of the workplace:<em> not enough time, missing data, ambiguous problems, SMEs who are slow to respond, collaborators who end up down rabbit holes, sick days, scope creep, difficult stakeholders, human error, pie charts</em>… and so on.</p><p>This is what <a href="https://cracked-it-book.com/"><em>Cracked It</em></a> from authors Bernard Garrette, Corey Phelps and Olivier Sibony is about. It provides a practical toolkit for anyone whose job is to influence decisions with the aid of data in large and complex organisations.</p><h3>Are you an Othello or a Hamlet?</h3><p><em>The core problem of problem-solving</em>, according to the authors, <em>is the tendency to think too fast and jump to solutions. We spend too little time and effort understanding a problem, believing instead we know all we need.</em> (p.4). This is demonstrated tragically by Shakespeare’s character Othello who kills his faithful wife under false interpretation of data.</p><p>On the other hand, we also risk not being able to come to any decisions at all, overwhelmed by all the possibilities and stuck in a state of analysis paralysis. This is exemplified by Hamlet whose endless hesitation leads to the collapse of his family.</p><p>The premise of the book is that, even with experience and expertise in a domain, it is easy to fall into the trap of acting too fast like Othello or acting too slow like Hamlet. In order to effectively balance the two, we need a systematic problem solving method, which the book presents as the <strong>4S framework</strong>.</p><blockquote><em>We need to know how to reason, in a generalisable way, to solve complex problems, without falling prey to analysis paralysis</em>. (p 11)</blockquote><h3>The 4S Method</h3><p><em>Cracked It</em> proposes the ‘4S Method’ as a system to follow for tackling any complex business problem, such as the question of whether to acquire another business or whether a new product idea will work in the market. The 4 ‘S’s, in order, are:</p><ol><li><strong>State</strong>: get a clear understanding of the problem we are setting out to solve and get alignment on this across stakeholders. A problem well stated is half solved.</li><li><strong>Structure</strong>: figure out how best to structure the next steps in the problem solving journey, picking from one of three approaches depending on the circumstances.</li><li><strong>Solve</strong>: gather data, perform analysis, generate candidate solutions for the problem.</li><li><strong>Sell</strong>: figure out how we’re going to persuade the decision makers on our recommended course of action.</li></ol><p>Though short in pages, the book is dense with sub-structures to use in each of these stages, and I found the real life (or disguised) examples to show the thinking processes in action effective. I have been using Notebook LM increasingly as a reading companion and I was also able to get a lot out of asking the AI to transpose the examples to situations that might occur in my own industry (for example for the ‘8 degrees of analysis’ in chapter 7) — this made the subject more immediately relatable to me and easier to absorb.</p><p>The ‘<strong>State</strong>’ step is all about problem definition, which early-career analysts tend to undervalue and is often a transformative part of their development journey from SQL technician to a valued business partner. The book proposes the ‘TOSCA’ framework for landing on good alignment here with stakeholders, going through the letters:</p><ul><li><strong>Trouble</strong>: What is the problem, why is it a problem and why now?</li><li><strong>Owner</strong>: Whose problem is this? This identity shapes the possible space of solutions.</li><li><strong>Success Criteria</strong>: What does success look like and by when?</li><li><strong>Constraints</strong>: What solutions can we not allow? What limitations might we face in the problem-solving process itself? (e.g. confidentiality, constraints on time to spend on this)</li><li><strong>Actors</strong>: Which other stakeholders will have influence on the solution? What do they want?</li></ul><p>I suspect TOSCA could work well as a template for analytics project planning as well. A disciplined consideration of the success criteria should help push back on the ‘laundry list’ attitude to stakeholder requests for data, while constraints could include specifics around data cleanliness/integrity. ‘Actors’ is a good one to consider upfront to identify peers and other stakeholders who it might be useful to reach out to early in the process to allow plenty of time for dialogue.</p><p>‘<strong>Structure</strong>’ and ‘<strong>Solve</strong>’ go into a lot of detail on specific approaches to take on using the time available to us for analysis. Whether it be through the hypothesis pyramid, issue tree or design thinking approach, a key takeaway here for analysts is that we should start with a structured plan upfront on the specific questions we want to validate with data and where we will go looking for this data. This is another potentially transformative shift in the career of analysts, going from someone who jumps right in and follows their nose to one who aligns on a plan based on formal logic upfront with stakeholders.</p><p>For me, the S that called out loudest was the ‘<strong>Sell</strong>’. This is about how we go about presenting our findings and our recommendations to decision makers in a way that is punchy and outcomes-focussed. Minto’s pyramid principle makes an appearance here and there is a wonderful example of how not to tell the story, and how that same story looks told in a more effective, top-down format. This is a struggle with many analysts even at senior levels, and I am inclined to share chapters 10 and 11 with all new joiners. Coud I staple it to their job contracts?</p><h3>Stray notes</h3><ul><li>I really appreciate how cognizant the book is of the very real consideration of personalities and politics in real world problem solving. For example: <em>It doesn’t take a Machiavelli of corporate politics to realise that the CEO’s vision isn’t a mere ‘hypothesis’ to be skeptically challenged. In most organisations, there is no such thing as a ‘hypothesis to be proven or disproven’ — proposals have proponents, precedents, and histories, and those who evaluate them know it. </em>(p.38)</li><li><em>Rapid advances in AI and robotics won’t make problem-solving skills any less relevant or important…</em> <em>it is likely to place a bigger premium on human problem solving.</em> (p10–11). I am inclined to agree that this is probably the best that we humans have to offer (more than the technical skills that distinguished my profession pre-AI) and I see the analytics profession moving more in this direction. The book was published in 2018 so it would be interesting to see what the authors think today now that the said ‘advances’ are in full swing and worrying everyone in a white collar.</li></ul><h3>Further reading</h3><ul><li>This book is often recommended alongside <a href="https://www.amazon.com.au/Bulletproof-Problem-Solving-Changes-Everything/dp/1119553024">Bulletproof Problem Solving</a>. The latter is a gentler introduction to problem solving frameworks and probably the better place to start before Cracked It.</li><li>For a more specific discussion on analytics workflows, I recommend <a href="https://medium.com/@juntaah/review-becoming-a-data-head-alex-j-gutman-jordan-goldmeier-d6854f18792b">Becoming a Data Head</a>, which is very readable at any level of experience in the field.</li><li>For an extensive library of general thinking tools, you have to know about <a href="https://fs.blog/">Farnam Street</a>.</li></ul><img src="https://medium.com/_/stat?event=post.clientViewed&referrerSource=full_rss&postId=56060aacfe39" width="1" height="1" alt="">]]></content:encoded>
        </item>
        <item>
            <title><![CDATA[How to Tackle the Weekend Quiz Like a Bayesian]]></title>
            <link>https://medium.com/data-science/how-to-tackle-the-weekend-quiz-like-a-bayesian-b5e035ba3746?source=rss-8cd5173a44bc------2</link>
            <guid isPermaLink="false">https://medium.com/p/b5e035ba3746</guid>
            <category><![CDATA[behavioural-science]]></category>
            <category><![CDATA[bayesian-statistics]]></category>
            <category><![CDATA[probability]]></category>
            <category><![CDATA[bias]]></category>
            <category><![CDATA[editors-pick]]></category>
            <dc:creator><![CDATA[Junta Sekimori]]></dc:creator>
            <pubDate>Mon, 28 Oct 2024 08:02:14 GMT</pubDate>
            <atom:updated>2024-10-28T08:02:14.541Z</atom:updated>
            <content:encoded><![CDATA[<h4>Do you know which of these is a malmsey? Can you make a good guess?</h4><p>A couple of weeks ago, this question came up in <a href="https://www.instagram.com/goodweekendquiz/p/DAt26iSPKeE">the Sydney Morning Herald Good Weekend quiz</a>:</p><blockquote>What is malmsey: a mild hangover, a witch’s curse or a fortified wine?</blockquote><p>Assuming we have no inkling of the answer, is there any way to make an informed guess in this situation? I think there is.</p><p>Feel free to have a think about it before reading on.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/512/0*2i73MfMD4_Y_9S6s" /><figcaption>A witch with a mild hangover from fortified wine, created using Gemini Imagen 3</figcaption></figure><h3>Is there really nothing we can bring to this question?</h3><p>Looking at this word, it <em>feels </em>like it could mean any of these options. The multiple choice, of course, is constructed to feel this way.</p><p>But there is a rational approach we can take here, which is to recognise that each of these options have different <em>base rates</em>. This is to say, forgetting about what is and isn’t a malmsey for a moment, we can sense that there probably aren’t as many names for hangovers as there are for witch’s curses, and there are bound to be even more names for all the different fortified wines out there.</p><p>To quantify this further:</p><ul><li>How many words for mild hangovers are there likely to be? <strong>Perhaps 1</strong>?</li><li>How many words for witch’s curses are there likely to be? I am no expert but I can already think of some synonyms so <strong>perhaps 10</strong>?</li><li>How many words for fortified wines are there likely to be? Again, not an expert but I can name a few (port, sherry…) and there are likely to be many more so <strong>perhaps 100</strong>?</li></ul><p>And so, with no other clues into which might be the correct answer, fortified wine would be a well reasoned guess. Based on my back-of-envelope estimates above, fortified wine would be x100 as likely to be correct as the mild hangover and x10 as likely as the witch’s curse.</p><p>Even if I am off with those quantities, I feel confident at least in this order of base rates so will go ahead and lock in fortified wine as my best guess.</p><p><a href="https://www.madeirawineanddine.com/malmsey/">Bingo</a>!</p><h3>Base rate neglect</h3><p>The reasoning may seem trivial but overlooking base rates when making judgements like this is one of the great human biases talked about by Kahneman and Tversky and many others since. Once we see it, we see it everywhere.</p><p>Consider the following brain teaser from Rolf Dobelli’s <a href="https://www.amazon.com.au/Art-Thinking-Clearly-Rolf-Dobelli/dp/0062219693"><em>The Art of Thinking Clearly</em></a>:</p><blockquote>Mark is a thin man from Germany with glasses who likes to listen to Mozart. Which is more likely? That Mark is A) a truck driver or (B) a professor of literature in Frankfurt?</blockquote><p>The temptation is to go with B based on the stereotype we associate with the description, but the more reasonable guess would be A because Germany has many, many more truck drivers than Frankfurt has literature professors.</p><p>The puzzle is a riff on Kahneman and Tversky’s librarian-farmer character portrait (see <a href="https://www2.psych.ubc.ca/~schaller/Psyc590Readings/TverskyKahneman1974.pdf"><em>Judgment under Uncertainty</em></a><em>)</em> which also provides the framing for the great <a href="https://www.youtube.com/watch?v=HZGCoVF3YvM">3B1B explainer on Bayes’ Theorem</a> where this kind of thinking process is mapped to the conditional and marginal probabilities (base rates) of the Bayes’ formula.</p><h3>Seeing the thinking traps</h3><p>The Bayesian framework helps us to more clearly see two common traps in probabilistic reasoning. In Kahneman and Tversky’s language, we could say it provides a tool for System II (‘slow’) thinking to override our impulsive and error-prone System I (‘fast’) thinking.</p><p>The first insight is that conditional probability of one thing given another p(A|B) is not the same as the probability of the reverse p(B|A), though in day-to-day life we are often tempted to make judgments as if they are the same.</p><p>In the Dobelli example, this is the difference of:</p><ul><li>P(👓|🧑‍🏫) — Probability that 👓) Mark is a thin man from Germany with glasses who likes to listen to Mozart <em>given that </em>🧑‍🏫) Mark is a literature professor in Frankfurt</li><li>P(🧑‍🏫|👓) — Probability that 🧑‍🏫) Mark is a literature professor in Frankfurt <em>given that </em>👓) Mark is a thin man from Germany with glasses who likes to listen to Mozart</li></ul><p>If stereotypes are to be believed, the P(👓|🧑‍🏫) above seems quite likely, whereas p(🧑‍🏫|👓) is unlikely because we would expect there to be many other people in Germany who fit the same description but aren’t literature professors.</p><p>The second insight is that these two conditional probabilities are related to each other, so knowing one can lead us to the other. What we need in order to connect the two are the individual base rates of A and B, and the scaling factor is in fact a simple ratio of the two base rates as follows:</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/242/0*MWTzX65TRYWHxr7A" /><figcaption>Image created by author</figcaption></figure><p>This is the Bayes’ formula.</p><h3>Bayesian Reasoning — step by step</h3><p>So how does this help us?</p><p>Outside of textbooks and toy examples, we wouldn’t expect to have all the numbers available to us to plug into Bayes’ formula but still it provides a useful framework for organising our knowns and unknowns and formalising a reasoned guess.</p><p>For example, for the Dobelli scenario, we might start with the following guesstimates:</p><ul><li>% of professors who wear glasses and fit the description: <strong>25%</strong> (1 in every 4)</li><li>% of people in Germany who are literature professors in Frankfurt: <strong>0.0002%</strong> (1 in every 500,000)</li><li>% of truck driver who wear glasses and fit the description: <strong>0.2%</strong> (1 in every 500)</li><li>% of people in Germany who are truck drivers: <strong>0.1%</strong> (1 in every 1,000)</li><li>% of the general population who wear glasses and fit the description:<strong> 0.2%</strong> (1 in every 500)</li><li>Population of Germany: ~<strong>85m</strong></li></ul><p>All these parameters are my estimates based on my personal worldview. Only the population of Germany is a data point I could look up, but these will help me to reason rationally about the Dobelli question.</p><p>The next step is to frame these in contingency tables, which show the relative frequencies of each of the events occurring, both together and individually. By starting with the total population and applying our percentage estimates, we can start to fill out two tables for the Frankfurt professors and truck drivers each fitting the description (for this section, feel free to also follow along in <a href="https://docs.google.com/spreadsheets/d/1DvPxzEFOeOEaKUc718MaqPCmS8PshAwZkRjPmYDNl-s">this spreadsheet</a>):</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/0*9-DzKsRIhE4-UKzU" /><figcaption>Image and resource created by author -see <a href="https://docs.google.com/spreadsheets/d/1aVMJUhzSCdj1s0IOWeO9-gVQ5NA84AV3RBwIrJr5msY">here</a> for original doc</figcaption></figure><p>The four white boxes represent the four ways in which the two events can occur:</p><ul><li>A and B</li><li>A but not B</li><li>B but not A</li><li>Neither A nor B</li></ul><p>The margins, shaded in grey, represent the total frequencies of each event regardless of overlap, which is just the sum of the rows and columns. Base rates come from these margins, which is why they are often referred to as <em>marginal probabilities.</em></p><p>Next, we can fill in the blanks like a sudoku by making all the rows and columns add up:</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/0*yf_57fqNetDh7OVg" /><figcaption>Image and resource created by author -see <a href="https://docs.google.com/spreadsheets/d/1aVMJUhzSCdj1s0IOWeO9-gVQ5NA84AV3RBwIrJr5msY">here</a> for original doc</figcaption></figure><p>And now, with our contingency tables complete, we have a full picture of our estimates around base rates and the likelihoods of the profiles matching the descriptions. All the conditional and marginal probabilities from the Bayes formula are now represented here and can be calculated as follows:</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/0*y5pVXxqtvbtYv0Vt" /><figcaption>Image and resource created by author -see <a href="https://docs.google.com/spreadsheets/d/1aVMJUhzSCdj1s0IOWeO9-gVQ5NA84AV3RBwIrJr5msY">here</a> for original doc</figcaption></figure><p>Back to the original question, the probability we are interested in is the third in the list above: the probability that they are a professor/truck driver given the description.</p><p>And, based on our parameter estimates, we see that truck drivers are x4 more likely to fit the bill than our professors (0.001 / 0.00025). This is in contrast to the reverse conditional where the description is more likely to fit the professor than a truck driver by a factor of x125 (0.25 / 0.002)!</p><h3>Back to Malmsey</h3><p>Now, looping back around to where we started with the malmsey example, hopefully the intuition is bedding in and the role of the base rates in making a guess is clear.</p><p>In terms of mapping the thinking to the Bayes formula, essentially, the thinking process would be to compare our degrees of belief of the following three scenarios:</p><ul><li>Probability (A the answer is <em>mild hangover</em> | B the word is <em>malmsey</em>)</li><li>Probability (A the answer is <em>witch’s curse</em> | B the word is <em>malmsey</em>)</li><li>Probability (A the answer is <em>fortified wine </em>| B the word is <em>malmsey</em>)</li></ul><p>Because in this case we have no inkling as to what malmsey could correspond to (this would be different if we had some etymological suspicions for example), we could say that B is <em>uninformative </em>and so to make any sort of reasoned guess, all we have to go by are the probabilities of A. In terms of the Bayes formula, we can see that the probability we are interested in scales with the base rate of A:</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/569/0*j9qebrGEBcOZqBAE" /><figcaption>Image created by author</figcaption></figure><p>For completeness, <a href="https://docs.google.com/spreadsheets/d/1DvPxzEFOeOEaKUc718MaqPCmS8PshAwZkRjPmYDNl-s/edit?gid=1825971281#gid=1825971281">here</a> is what it might look like to tabulate our degrees of belief in the style of the contingency tables from the Dobelli example. Because B is uninformative, we give 50:50 odds for the word malmsey matching any other word or concept. This is overkill and hardly necessary once we recognise that we can simply scale our belief in the answer with the base rates, but it’s there to show the Bayesian framework still fits together for this more abstract problem.</p><h3>Base rate neglect in Hypothesis (A/B) Testing</h3><p>I previously <a href="https://towardsdatascience.com/how-to-intuit-the-prosecutors-fallacy-and-run-better-hypothesis-tests-2d8561bfc5a6">wrote on the topic of the prosecutor’s fallacy</a> (a form of base rate neglect) which gives other examples on base rate neglect and implications for analytics practitioners.</p><p>It is worth making the connection again here that in conventional A/B testing methods, people often confuse the probability they get of seeing the test results with the probability of the hypothesis itself being true. Much has been written about p-values and their pitfalls (see, for example, <a href="https://sixsigmadsi.com/wp-content/uploads/2020/10/A-Dirty-Dozen-Twelve-P-Value-Misconceptions.pdf">A Dirty Dozen: Twelve P-Value Misconceptions</a>), but this is another place where the Bayesian mindset helps to clarify our reasoning and where it helps to be alert to the concept of base rate neglect, which in this case is our confidence in the hypothesis being true in the first place (our <em>priors</em>).</p><p>I encourage you to read <a href="https://towardsdatascience.com/how-to-intuit-the-prosecutors-fallacy-and-run-better-hypothesis-tests-2d8561bfc5a6">the article</a> to get a better intuition for this.</p><h3>Takeaways</h3><ul><li>Concepts covered: base rate neglect, conditional vs marginal probabilities, Bayes’ formula, contingency tables.</li><li>Be careful not to equate p(A|B) with p(B|A) in day-to-day judgement of likelihoods.</li><li>Consider base rates when making a judgement of whether a new observation validates your hypothesis.</li><li>TIL: Malmsey is a fortified wine from the island of Madeira. In Shakespeare’s Richard III, George Plantagenet the Duke of Clarence drowns in a vat of malmsey.</li></ul><h3>Further reading</h3><ul><li><a href="https://www.amazon.com.au/Art-Thinking-Clearly-Rolf-Dobelli/dp/0062219693">The Art of Thinking Clearly by Rolf Dobelli</a> where this article takes the professor-truck driver puzzle from provides a great, readable collection of thinking traps in day-to-day life. Alongside the chapter on base rate neglect (ch 28), I enjoyed the framings used for regression to the mean (ch 19), exponential growth (ch 34), false causality (ch 37) and many more. Each chapter is succinct at 2–3 pages and the book works well as an ongoing reference manual for common biases and fallacies.</li><li><a href="https://www.amazon.com.au/Undoing-Project-Michael-Lewis/dp/0141983043">The Undoing Project by Michael Lewis</a> (author of The Big Short) is the story of Kahneman, Tversky and the development of behavioural economics. It has all the best bits from Thinking Fast and Slow and is a page turner that could easily become a movie adaptation.</li><li><a href="https://www.amazon.com.au/Learning-Data-Statistics-David-Spiegelhalter/dp/0241258766">The Art of Statistics by David Spiegelhalter</a> has great, layman-friendly chapters on Bayesian statistics.</li><li><a href="https://towardsdatascience.com/how-to-intuit-the-prosecutors-fallacy-and-run-better-hypothesis-tests-2d8561bfc5a6">How to Intuit the Prosecutor’s Fallacy (and run better hypothesis tests)</a> is a previous article I wrote on a similar topic when I was grappling with the proper definition of p values.</li><li>For getting an intuition for the Bayes formula, I recommend watching the Stat Quest videos on <a href="https://www.youtube.com/watch?v=_IgyaD7vOOA">1. Conditional Probability</a> and <a href="https://www.youtube.com/watch?v=9wCnvr7Xw4E">2. Bayes’ Theorem</a>.</li></ul><img src="https://medium.com/_/stat?event=post.clientViewed&referrerSource=full_rss&postId=b5e035ba3746" width="1" height="1" alt=""><hr><p><a href="https://medium.com/data-science/how-to-tackle-the-weekend-quiz-like-a-bayesian-b5e035ba3746">How to Tackle the Weekend Quiz Like a Bayesian</a> was originally published in <a href="https://medium.com/data-science">TDS Archive</a> on Medium, where people are continuing the conversation by highlighting and responding to this story.</p>]]></content:encoded>
        </item>
        <item>
            <title><![CDATA[Review: Becoming a Data Head, Alex J. Gutman & Jordan Goldmeier]]></title>
            <link>https://medium.com/@juntaah/review-becoming-a-data-head-alex-j-gutman-jordan-goldmeier-d6854f18792b?source=rss-8cd5173a44bc------2</link>
            <guid isPermaLink="false">https://medium.com/p/d6854f18792b</guid>
            <category><![CDATA[analytics]]></category>
            <category><![CDATA[leadership]]></category>
            <category><![CDATA[management]]></category>
            <category><![CDATA[data-science]]></category>
            <category><![CDATA[books]]></category>
            <dc:creator><![CDATA[Junta Sekimori]]></dc:creator>
            <pubDate>Thu, 22 Feb 2024 13:50:45 GMT</pubDate>
            <atom:updated>2024-02-25T03:15:06.340Z</atom:updated>
            <content:encoded><![CDATA[<h4>A readable and practical guide for navigating organisational challenges in the real-world application of data analytics.</h4><figure><img alt="" src="https://cdn-images-1.medium.com/max/700/0*wW-JUieyklQJ6Bdt.jpg" /><figcaption><a href="https://www.amazon.com.au/Becoming-Data-Head-Understand-Statistics/dp/1119741742">Link to book</a>.</figcaption></figure><p>Data analytics in theory is different to data analytics in practice. The tools we learn in university or online turn out to not be so simple to apply in the workplace even with the greatest technical competence.</p><p>Learning to practise our craft amidst all the moving parts of a real-life organisation is a big part of the journey we take in our careers, from our first forays in summarising data to managing end-to-end data analytics projects with all the twists and turns that these always entail. <em>Becoming a Data Head</em> is an A-Z guide of these twists and turns, and how to guide your teams/stakeholders through them.</p><h4>Simple explanations</h4><p>As a manager of an analytics team, people look to me to give simple answers to complex questions. Not just in terms of explaining insight or the technicalities of an analysis/model, but also to diabolical questions such as <em>how long will it take</em>?</p><p>Perhaps the greatest expectation gap I regularly have to overcome is between the idea that data = science = hard truths, and the reality that working with data is fundamentally the art of working with uncertainty.</p><p>Uncertainty around the trustworthiness of the data we are working with. The uncertainty from natural random variation in the observations. The uncertainty of what is correlation and what is causal. The uncertainty of what the stakeholders are <em>really </em>looking for (and the ensuing madness of trying to read between the lines a bit too much).</p><p>The list goes on and we find ourselves forever resisting to respond with a million terms and conditions (climaxing in <em>what even is probability? </em>which I am 99.9% sure is a debate my stakeholders will not enjoy having during their 9 to 5).</p><p>So this is the book about navigating organisational challenges in the pursuit of data excellence. How can we manage all these uncertainties of working with real-world data in fast moving organisations? What is it to manage a programme of work in data analytics/science and what does it take to be successful beyond the technical skills?</p><h4>Tools of management</h4><p>If there are books about filling us with optimism about the tools of data science, this is the book of realism that equips us with the tools of management to actually make data count in the workplace. With every passing year in these trenches, I find myself appreciating with increasing conviction that this is a skill to learn, develop and constantly sharpen, just as much as ‘hard’ skills like SQL or how to construct a regression model.</p><p>It is at once an entry level and senior level book. Concepts are taught simply as if for beginners (I love the writing style and highly recommend it for this), but it would take people with a few scars to appreciate the pitfalls and the tips on overcoming organisational difficulties. The title of the book evokes this well, addressing both the general data enthusiast — i.e. in the same sense as <em>petrolhead</em> — and the ‘Head Of’ in an organisation.</p><p>The book starts with the premise of why problems occur in the first place. These are given as follows (paraphrased):</p><ol><li>Problem-solving with data is frankly a very complex process. The inference or predictions will not always hit the mark, even with the most competent team in the world.</li><li>Some analysts and decision makers have stopped thinking critically about the data.</li><li>There is a communication gap between decision makers who don’t speak data and the data workers who can’t seem to explain things simply.</li></ol><p>Thinking critically about data in particular forms the backbone of the ensuing 200 or so pages and every new explanation of a data science concept is typically followed by commentary on what to watch out for when seeing it in real life. Key concepts in general statistical thinking are covered such as the traps of conditional probability, confidence intervals and correlation vs causation, as are applied data science concepts starting with an intuition for supervised vs unsupervised learning and ending in deep learning.</p><h4>Data therapy</h4><p>There is probably not much new here for anyone who has spent time in the field (the anecdotes are also quite well worn: lessons from the challenger disaster, smoking causes cancer etc) but the articulation is clear and I see myself using this as a reference for when I have to explain concepts or put together material for teaching.</p><p>For me, the appendix-like last part of the book that spans just 20 pages, was alone worth the price of the book. Comprising the two chapters ‘Watch Out for Pitfalls’ and ‘Know the People and Personalities’, this section neatly summarises the day to day problems that come up during data projects in the workplace. I found it agonisingly relatable and it felt like therapy reading some of these workplace vignettes.</p><p>The pitfalls here include classic challenges in interpretation such as survivorship bias and simpson’s paradox, and also lapses of reasoning such as p-hacking and confusing correlation for causation. Pitfalls are not just data pitfalls and there is a third list of <em>project</em> pitfalls such as ‘not adapting the question once it’s already failed’ and ‘overkill’.</p><p>The ‘Seven Scenes of Communication Breakdown’ was triggering. My favourite of the seven was <em>The Telephone Game</em> in which an unvetted figure from the early stages of analysis is taken out of context and widely circulated around the organisation by excited stakeholders.</p><blockquote><em>The data scientist realises this factoid came from a casual coffee meeting and was never meant to be presented without vetting. At this point, the factoid has been repeated so many times within the company, it’s said as if it were self-evidently true. She wonders how much she could push back on the company using it — and if it’s even worth doing so.</em></blockquote><blockquote><em>P.206</em></blockquote><h4>Stray notes</h4><ul><li>I liked the framing of the “data science industrial complex” where new tools and techniques are constantly overhyped and oversold to organisations. (xxvii)</li><li>I had an aha moment in the hypothesis testing section where a question is posed to the reader about which way around to set the null hypothesis and alternative hypothesis. Does it matter? It does, because <em>the burden of proof</em> should be on the observed data to reject the status quo. The hypothesis statements should be framed in this way, or we risk an error of reasoning. (p.92–93)</li><li>There is a section on text analytics that goes into sentiment classification, n-grams and word embeddings. This was only published in 2021 and already it is beginning to feel quaint amidst all the recent LLM explosion. (p.151–170)</li></ul><h4>Further reading</h4><ul><li>For a more in-depth but no less accessible primer on statistical thinking, I recommend <a href="https://medium.com/@juntaah/review-the-art-of-statistics-david-spiegelhalter-ed366f4102ea">The Art of Statistics</a>.</li><li>The book recommends two other books for further reading. One of them is <a href="https://www.amazon.com.au/Data-Smart-Science-Transform-Information/dp/111993138X/">Data Smart</a>. I read and loved the first version of the book by John W. Foreman which teaches the intuition of data science methods (logistic regression, bagging/boosting, ensemble methods etc) using Excel formulae. I have just noticed on Amazon the second edition of the book has just been published by one of the authors of this book!</li><li>For workplace project/stakeholder management, I recently got a lot out of <a href="https://www.amazon.com.au/Bulletproof-Problem-Solving-Changes-Everything/dp/1119553024">Bulletproof Problem Solving</a>, and now directly use its frameworks at work.</li></ul><img src="https://medium.com/_/stat?event=post.clientViewed&referrerSource=full_rss&postId=d6854f18792b" width="1" height="1" alt="">]]></content:encoded>
        </item>
        <item>
            <title><![CDATA[How to Intuit the Prosecutor’s Fallacy (and Run Better Hypothesis Tests)]]></title>
            <link>https://medium.com/data-science/how-to-intuit-the-prosecutors-fallacy-and-run-better-hypothesis-tests-2d8561bfc5a6?source=rss-8cd5173a44bc------2</link>
            <guid isPermaLink="false">https://medium.com/p/2d8561bfc5a6</guid>
            <category><![CDATA[ab-testing]]></category>
            <category><![CDATA[data-analysis]]></category>
            <category><![CDATA[data-science]]></category>
            <category><![CDATA[editors-pick]]></category>
            <category><![CDATA[statistics]]></category>
            <dc:creator><![CDATA[Junta Sekimori]]></dc:creator>
            <pubDate>Wed, 17 Aug 2022 07:50:54 GMT</pubDate>
            <atom:updated>2022-08-17T10:54:49.017Z</atom:updated>
            <content:encoded><![CDATA[<h4>For data scientists and analysts in business</h4><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/0*zeEon7qBpUUl0Wov" /><figcaption>Photo by <a href="https://unsplash.com/@tingeyinjurylawfirm?utm_source=medium&amp;utm_medium=referral">Tingey Injury Law Firm</a> on <a href="https://unsplash.com/?utm_source=medium&amp;utm_medium=referral">Unsplash</a></figcaption></figure><p><em>There is only a one in a million chance that the accused would match the DNA found at the crime scene</em>. So the accused is guilty beyond reasonable doubt. Sound ok? It isn’t!</p><p>Based on this evidence alone, the accused may have a good chance of being innocent and this is the logical trap described by the unintuitive but wonderfully illuminating <strong>Prosecutor’s Fallacy</strong>.</p><p>When we understand this, we start to see it everywhere in advertising, the media and people’s day-to-day decision making. It also underpins <a href="https://fivethirtyeight.com/features/not-even-scientists-can-easily-explain-p-values/">a common difficulty</a> in correctly interpreting p-values, and getting a better intuition for this brings a lot more clarity to how we can approach hypothesis testing in business settings.</p><p>One more example to get the juices going. A medical test you have taken for a serious disease has come back positive. You feel a rush of panic as you recall that the test has an accuracy rate of 95%. And yet your real odds may be a lot better than that — perhaps a 10% chance that you actually have it. How can that be?</p><h4>Learning objectives</h4><p>The first objective of this article is to give a general introduction to the Prosecutor’s Fallacy and explain it in plain language, targeting the reasons that make it difficult to intuit. If you haven’t come across it or are hazy on the concept, it’s a great tool for life whatever path you walk, and I hope you will find learning about it as enjoyable as I did.</p><p>The second part of the article will go into how the Prosecutor’s Fallacy plays into the interpretation of <strong>p-values</strong> in the use of <strong>hypothesis testing</strong> and is targeted at the practising data professional. If you struggle to explain why p-values are not the probability of your null hypothesis being false but rather <em>the probability of you getting results as extreme as those observed given the null hypothesis</em>, it’s worth understanding — read on.</p><h3>Understanding the Prosecutor’s Fallacy</h3><h4>Exhibit A</h4><p>So the chance that the accused would match the DNA from the crime scene is one in a million. This is a highly accurate DNA test, and there has been a positive match. And yet, however compelling this evidence may seem, this alone does not incriminate the accused. Not because there is still a small amount of uncertainty in this probability (there is never absolute certainty), but because <em>the probability of the evidence given innocence</em> is not the same as the <em>probability of innocence given the evidence</em>.</p><p>To understand what this means, let’s first suppose we are looking for the offender in a city of 10 million people (plus the one offender). And there is a 100% certainty that the true offender’s DNA will match the DNA at the crime scene.</p><p>Putting these numbers in a table, we get this:</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/717/1*yHDNsLHOObBB_B7JYkToGw.png" /><figcaption>Image by author</figcaption></figure><p>Now, we have been told that there is only a one in a million chance that an innocent person would get the DNA match. So if we were to test all 10 million people in the city, we would expect 10 of those to match.</p><p>Adding this into the table:</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/717/1*WbA7u8qOb9gIt8xGUEaedQ.png" /><figcaption>Image by author</figcaption></figure><p>Now look at all the innocent people who would match the DNA in this scenario. In fact, the number of innocent matches outweigh the correct guilty match ten to one! That is 10 false positives to the one true positive.</p><p>Expressed as a percentage, the probability of guilt given the evidence across the whole city then is 1 / (1 + 10) = 9.1%. On the basis of this DNA evidence alone, it now looks like the accused has a 1 – 9.1% = 90.9% chance of innocence, which is in stark contrast to the 0.0001% innocence that was initially claimed by the prosecutor.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/717/1*uW7rIk9spYkeK4HCBDD6RA.png" /><figcaption>Image by author</figcaption></figure><p>The implication of the original claim was that, since only one in a million innocent people would match the DNA and the accused matched the DNA, the probability of their guilt is 1 – 0.000001, a near certainty. But this logic is flawed and particularly problematic when the population the probability applies to is large; the conclusion we’d draw from the data could dramatically flip like in this example. The probability of innocence given the DNA match must not be confused by the probability of a DNA match given innocence.</p><p>So is that to say we can’t do anything with DNA evidence in criminal trials? We can, but key here is to understand that there needs to be other reasons to believe in the guilt of the accused. If we relied just on the DNA match, we expose ourselves to spurious matches, but if we already had other suspicions about someone, then this additional evidence would be extremely compelling. We could also use DNA to narrow down suspects to help focus the investigation (as per <a href="https://www.youtube.com/watch?v=KT18KJouHWg">the hunt for the Golden State killer</a>), but it is not conclusive, and this has important parallels to the use of p-values in hypothesis testing which we will explore in the second part.</p><p>An extension to the Prosecutor’s Fallacy, worth quickly mentioning here, is the <strong>Defender’s Fallacy</strong>, in which it is incorrectly argued that a piece of evidence incriminating the defendant should be discarded because it could be matched against many others. This was used in the trial of OJ Simpson where the defending lawyers argued that the blood samples from the scene matching OJ’s (1 in 400 match rate) could also be matched against thousands of others across the LA population and so was not useful evidence. This is a misrepresentation of the Prosecutor’s Fallacy and the evidence is useful because it is considered in combination with other evidence, despite not being conclusive on its own.</p><h4>Exhibit B</h4><p>It was the news you had been dreading. You are healthy but have been screened for a rare disease and the ‘95% accurate’ test you took has come back positive. You feel numb. But should you?</p><p>Here we first run into the question of what ‘accuracy’ means and <a href="https://www.pathology.health.nsw.gov.au/covid-19-testing/sensitivity-and-specificity">in medical testing</a>, the terms <strong><em>sensitivity</em></strong> and <strong><em>specificity</em></strong> are used to definite two different aspects of accuracy:</p><ul><li><strong>Sensitivity</strong>: the likelihood that people with the disease will be correctly identified as having the disease (the true positive rate)</li><li><strong>Specificity</strong>: the likelihood that people without the disease will be correctly identified as not having the disease (the true negative rate)</li></ul><p>For this example we will assume the 95% accuracy applies to both the <em>sensitivity</em> and <em>specificity</em>. And we will assume the disease affects 1 in 1,000 of the population.</p><p>In a population of 10 million therefore, we would expect to see 10,000 people with the disease. Putting these numbers in a table gives us this picture:</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/731/1*LtbV0kJ_TiBe6CkHltjSWw.png" /><figcaption>Image by author</figcaption></figure><p>Another way of representing this which might be a little more intuitive is using a decision tree:</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/733/1*Fau7iaoGVY0HGZ35RCmqdw.png" /><figcaption>Image by author</figcaption></figure><p>With these numbers, the odds that you have the disease given you have just had a positive test (and have no other reasons to believe you have the disease) works out to be 9,500 / (9,500 + 499,500) = 1.87%.</p><p>Even if the disease was x10 more prevalent in the population, your likelihood of having it after getting a positive test goes up to just 16.1%. On top of this, increase the specificity of the test to 99%, and still your odds are 49%. (For a bit of extra context, in Australia, COVID at-home rapid antigen test kits <a href="https://www.tga.gov.au/covid-19-rapid-antigen-self-tests-are-approved-australia">have to have been certified</a> for a sensitivity of &gt;80% and a specificity of &gt;98%).</p><p>The initial 95% ‘accuracy’ measure did not give you enough information on the actual probability of having the disease because you were missing information on the prevalence of the disease in the population, and therefore the prior likelihood of you having the disease.</p><p>The probability of you getting a positive test result given you have the disease (95%) is different to the probability that you have the disease given the test result, and to make any judgement on your likelihood of having the disease, you need some sense of how likely you are to have it in the first place, regardless of the test result (great commentary from the ABC using real-world COVID scenarios <a href="https://www.abc.net.au/news/2022-01-27/how-accurate-rapid-antigen-test-more-looking-for-lines/100782662">here</a>).</p><p>Prior probabilities and the idea of updating probabilities as new information (e.g. COVID test results) comes in is formalised in the Bayesian framework, and the famous Bayes formula provides a shortcut for working out what we have just been illustrating through contingency tables. This is a topic for another day.</p><h3>How the prosecutor’s fallacy comes into hypothesis testing</h3><p>From this point, I assume you have some background in hypothesis testing and have working knowledge of concepts such as the test statistic, the null hypothesis and p-values. <a href="https://towardsdatascience.com/p-values-explained-by-data-scientist-f40a746cfc8">This</a> might be a good place to go next if not and you’re interested.</p><p>The focus of this section will be on getting a better intuition for p-values and specifically the idea that the p-value is <strong>not</strong> the probability that the null hypothesis is true; p-value of 0.05 is <strong>not </strong>a 5% probability that the null hypothesis is true, nor a 95% probability that the alternative hypothesis is true. Rather, it is <em>the probability of getting the result observed (or more extreme), given the null hypothesis</em>, and this difference is difficult to intuit, let alone help others make sense of in practice when organising tests in the real world.</p><h4>Does it matter?</h4><p>First of all, does this difference in wording even really matter? For all intents and purposes, isn’t it the same thing? It isn’t, and the reason comes from the same distinctions that make the Prosecutor’s Fallacy a trap with serious practical consequences.</p><p>The danger with not understanding the difference is that we may draw unfounded conclusions from our test results, much like the prosecutor made a flawed claim against the defendant’s innocence.</p><p>Concretely, the problem comes when we think it’s ok to:</p><ul><li>Repeat tests across multiple segments until we land a p-value under 0.05</li><li>Split our initial test results into subsets and calling out any that come up with a p-value under 0.05 as meaningful</li></ul><h4>Exhibit C</h4><p>Imagine you are a supermarket testing an ad for Coke on your website and have split your online customers so that half see the ad and half don’t in an A/B test. You carefully calculate the sample sizes you need and let the data accumulate, but at the conclusion of the test you are disappointed to see that there is no significant difference in Coke sales between the two groups.</p><p>But, you think, what if it resonated with certain groups of customers? So you go exploring different breakdowns and start to notice results with p &lt; 0.05 here and there: customers who live in Newcastle, retirees who have a leaning towards value products, people who buy a lot of cheese…</p><p>The marketers get excited. Maybe we have something here. Yes, it makes sense that people from Newcastle would respond well to Coke ads because we are currently running a large billboard ad in the city centre there and it is already front of mind for the customers. Let’s double down and keep targeting this group online.</p><p>This is faulty reasoning akin to the prosecutor randomly picking people from the population until there is a positive DNA match and claiming guilt despite the 9.1% real chance of it. The fact that there seems to be evidence in support of this (the billboard) is confirmation bias; if we go looking for justifications, we often find them, but just because they are plausible in hindsight don’t make them relevant. Be careful not to post-rationalise.</p><p>To unpack this, we will use the contingency tables again. Taking again the numbers from our first example, the shaded section is essentially what p-values represent.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/526/1*6bzu15Y-rOBwab0QQICppA.png" /><figcaption>Image by author</figcaption></figure><p>But really what we’re interested in is:</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/526/1*wXLQC3BuklNV7HgWd7_deg.png" /><figcaption>Image by author</figcaption></figure><p>So why can’t we just measure this? Simply because we don’t have a way to measure the number for that bottom ‘guilty’ box when we are conducting hypothesis tests. The best we can do is essentially fill out the top row by collecting sample data and make an inference.</p><p>Transposing this to our coke ad A/B test example we get something that might look like the following if we did indeed observe a ‘significant’ difference from the ad with a p-value of 0.05:</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/719/1*4IkZouFy9AER7SxX1dJ_KQ.png" /><figcaption>Image by author</figcaption></figure><p>This gets confusing with the double negatives but the top row is saying that we would only have got the sales result we did (or more extreme) only 5 out of 100 times if the coke ad really made no difference to sales. It is a very surprising result, given the null hypothesis.</p><p>But crucially, what this doesn’t say is anything about whether the coke ad actually makes a difference to sales (the true effect under the alternative hypothesis). In order to get a position on that, we would need a number in the bottom left box and this is not something we can collect data on through A/B testing or any other way.</p><p>To help us conceptualise why the 0.05 p-value bears no direct relationship with the effectiveness of the coke ad, let’s try putting in some hypothetical numbers.</p><p>What could go in here? Let’s start with ‘1’. It’s an arbitrary number that only has meaning in relation to the ‘5’ in the box above but we’re following the same logic that we used in our earlier examples to illustrate the Prosecutor’s Fallacy.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/719/1*f-O15SHzXSNBQ97Ham1ZIQ.png" /><figcaption>Image by author</figcaption></figure><p>So now, despite the p-value of 0.05, we get a probability of the coke ad making a difference of just 1 / (1+5) = 17%.</p><p>Now let’s try putting in a bigger number; ‘45’, let’s say. This gives us a probability that the ad makes a difference of 45 / (45+5) = 90%. Same p-value, two starkly opposed conclusions we could draw from the data.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/719/1*viJJLoEbqFHxi0qU5mVgvg.png" /><figcaption>Image by author</figcaption></figure><p>But what’s the point of throwing around these hypothetical numbers that aren’t based on real, measurable data? This is the crux of this discussion and the key to the better practice of hypothesis testing in the real world. <em>The numbers you put in here represent your prior confidence in the hypothesis.</em></p><p>The higher the number, the higher your confidence. And what we have demonstrated by plugging in different numbers here is that a p-value of 0.05 has no relation to the probability of our hypothesis being true. If the test is a moonshot to begin with, you would still have cause to be sceptical even with favourable p-values. <strong>Prior conviction matters in hypothesis testing</strong>.</p><p>As touched on earlier, the Bayesian model of thinking provides an approach for handling this prior conviction mathematically, but the frequentist model of p-values and confidence intervals are the de facto standard for A/B testing in the business world today, and this trap in reasoning is too often overlooked. P-values don’t replace the need for human reasoning and we have to start with some prior conviction in the hypothesis for the test results and p-values to be useful. If ever this becomes unintuitive, think back to the prosecutor’s example.</p><h3>Putting it all together</h3><ol><li>A p-value tells you little about your hypothesis being true. Don’t draw conclusions purely on if a test passed a certain p-value threshold.</li><li>A p-value is the probability of observing the results or more extreme, given the null hypothesis. This is different to the probability of the null hypothesis given the evidence. The Prosecutor’s Fallacy illustrates why this distinction matters.</li><li>A p-value from an A/B test helps to validate our hypothesis, if we had good reason to believe in it in the first place.</li><li>Don’t go looking for attractive p-values and reasoning backwards from them. The best thing to do if you spot an unexpectedly low p-value for a segment is to validate it with a new test specific to that segment.</li></ol><h3>Using p-values in business decision-making</h3><p>When practising hypothesis testing with p-values in the business world, be clear-minded about their correct use and be prepared to explain it, but don’t feel you need to fiercely oppose every misuse of the p-value to be a good adviser.</p><p>We are working in environments where decisions need to be made fast and data is imperfect, so don’t get hung up if the decision-makers need to make decisions on their hunch from time to time based on unrigorous data; after all, confidence is part of the equation.</p><p>So even in situations where you know the right thing to do is to rerun a test, if the cost of doing so outweighs the cost of going ahead with the decision, then it might not make good practical sense. Be clear on your position but no need to fight for rigour on every decision.</p><h3>Further reading</h3><p><a href="https://medium.com/@juntaah/review-the-art-of-statistics-david-spiegelhalter-ed366f4102ea">The Art of Statistics</a> describes p-values as ‘weak evidence’ (i.e. a factor in decision making but far from the only factor) and provides some great commentary on how we might layer other evidence on top to build a strong case.</p><p>In courts of law, for example, different pieces of evidence can be itemised and each given different likelihood ratios (probability of evidence if accused is guilty / probability of evidence if accused is innocent), with a possibility of multiplying them all together to quantify the strength of a case, though courts in the UK stop short from doing so as that is the role of the jury.</p><p>The chapters on <em>Answering Questions and Claiming Discoveries </em>and <em>Learning from Experience the Bayesian Way </em>are full of great real-world examples such as this and the different schools of thought for weighing up evidence.</p><p>I took some inspiration also from the lecture series <a href="https://www.callingbullshit.org/">Calling Bullshit </a>and in particular the lecture on ‘<a href="https://www.youtube.com/watch?v=eesUdFlYMh8">P Values and the Prosecutor’s Fallacy</a>’.</p><img src="https://medium.com/_/stat?event=post.clientViewed&referrerSource=full_rss&postId=2d8561bfc5a6" width="1" height="1" alt=""><hr><p><a href="https://medium.com/data-science/how-to-intuit-the-prosecutors-fallacy-and-run-better-hypothesis-tests-2d8561bfc5a6">How to Intuit the Prosecutor’s Fallacy (and Run Better Hypothesis Tests)</a> was originally published in <a href="https://medium.com/data-science">TDS Archive</a> on Medium, where people are continuing the conversation by highlighting and responding to this story.</p>]]></content:encoded>
        </item>
        <item>
            <title><![CDATA[Review: The Art of Statistics, David Spiegelhalter]]></title>
            <link>https://medium.com/@juntaah/review-the-art-of-statistics-david-spiegelhalter-ed366f4102ea?source=rss-8cd5173a44bc------2</link>
            <guid isPermaLink="false">https://medium.com/p/ed366f4102ea</guid>
            <category><![CDATA[book-review]]></category>
            <category><![CDATA[analytics]]></category>
            <category><![CDATA[statistics]]></category>
            <category><![CDATA[books]]></category>
            <category><![CDATA[data-science]]></category>
            <dc:creator><![CDATA[Junta Sekimori]]></dc:creator>
            <pubDate>Sun, 31 Jul 2022 02:00:22 GMT</pubDate>
            <atom:updated>2026-04-10T06:40:50.822Z</atom:updated>
            <content:encoded><![CDATA[<h4>The best primer on statistics from today’s most joyful spokesperson on the subject</h4><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*w5-IbyX258JIsAl1_fh3IA.jpeg" /><figcaption>Affiliate link to purchase the book <a href="https://bookshop.org/a/122361/9781541675704">here</a> (thank you!)</figcaption></figure><h3>In brief</h3><p>An excellent guide for gaining a better intuition for the statistical concepts in use in the real world today, with great concrete examples. Big emphasis on process and where things go wrong in practice. An essential read for practitioners to help us better articulate how and why we use statistics. Should feel readable to a general interest or beginner audience but a better place to start might be <a href="https://bookshop.org/a/122361/9780393347777"><em>Naked Statistics</em></a> which has a more pop science feel.</p><h3><strong>Extended ramblings</strong></h3><p>In my work in business analytics I find myself constantly having to come back to topics like the problem of HARKing or the difference between statistical significance and statistical power. They are concepts that enable the work I do but when it comes to explaining these simply, I would often fumble for the right words and question my own understanding of them. I would revisit the material and feel confident again for a while, but they are tricky to articulate well without regular rehearsal. What makes statistical decision-making work is unintuitive, but it does work, and that is what makes it so rewarding to learn and teach.</p><p><em>The Art of Statistics</em> is a general primer on statistics, starting with mean, mode, medium and how distributions work. It is written for the general reader with no prior knowledge of statistics and its emphasis is on the practice of statistics in the real world. It covers a lot of ground, and touches on concepts such as the Bonferroni correction and ROC curves which seem ambitious for a general audience but are explained very well in simple language. As a practitioner with a constant need to educate and motivate non-specialists in my workplace (and myself!), I devoured this and found it a very useful reference for good explanations and examples.</p><p>I had a few a-ha moments. The first was around the idea of ‘population’. It has never sat easy with me that even in situations when we seem to have all the data, we would still consider that data to be a sample of a population. But the ‘population’ doesn’t need to actually exist in the world for it to be applied in statistics and he explains this beautifully, acknowledging that this is a more difficult idea than textbooks typically make out.</p><p>To use the example in the book, consider the survival rates of all children across a number of hospitals; even though we have collected all the data from all the operations performed, statistically we would still treat these as samples when determining whether there are significant differences in survival rates between the hospitals.</p><blockquote><em>But what is the study population? We have data on all the children and all the hospitals, and so there is no larger group from which they have been sampled. Although the idea of a population is introduced rather casually into statistics courses, this example shows it’s a tricky and sophisticated idea that is worth exploring in some detail. — p.91</em></blockquote><p>He goes on to list three kinds of populations and in particular the concept of a <em>metaphorical </em>population (in contrast to a <em>literal </em>population), which he describes as an imaginary space of possibilities that survival rates (or murder rates in a city or test results in a classroom) are drawn from. Essentially, it is what our dataset would look like if we had an infinity of data points.</p><blockquote><em>Personally, I rather like acting as if all that occurs around us is the result of some random pick from all the possible things that could happen. It is up to us whether we choose to believe it is truly chance, whether it is the will of a god or gods, or any other theory of causation: it makes no difference to the mathematics. This is just one of the mind-stretching requirements for learning from data. — p.93</em></blockquote><p>To touch on some other personal highlights:</p><ul><li>I found I got a better intuition for the prosecutor’s fallacy (the difference between ‘<em>if the accused is innocent, there is only a 1 in a billion chance that they would match the DNA found</em>’ and ‘<em>given the DNA evidence, there is only a 1 in a billion chance that the accused is innocent</em>’ — p.216)</li><li>This had the best real-life example and explanation of a poisson distribution I’ve come across (frequency of daily recorded homicide incidents in the UK — p.223–226). It’s amazing really what the poisson distribution predicts with so few inputs and he goes onto reflect on <em>the astonishing predictability of overall patterns made up of individually unpredictable events.</em></li><li>I found super compelling the distinction between <em>aleatory</em> and <em>epistemic</em> uncertainty and their philosophical links to frequentist and Bayesian statistics respectively — p.240).</li></ul><p>Again, the book’s focus is on statistics in practice and it was engaging to unpack questions such as ‘<em>How many trees are there in the world?</em>’, or ‘<em>Is prayer effective?</em>’ with the use of statistical techniques. Each of the chapters are supported by questions like these to guide the thinking and it was clarifying to not only see the suggestions on approach but also where it can go wrong in practice (how relatable too).</p><p>One difficulty with the practice of statistics is that there are many subjective choices to be made, which leads him to the admission that ‘<em>the formal basis for learning from data is a bit of a mess[…] it is no wonder that mathematicians tend to dislike teaching statistics. (p.305)</em>’. P-value thresholds are an easy example of this but more fundamentally the very definition of ‘probability’ is still debated. But the differences in these opinions tend not to matter where the final conclusions are concerned:</p><blockquote><em>My personal view is that, while they may well disagree about the fundamentals of their subject, reasonable statisticians will generally come to similar conclusions. The problems that arise in statistical science do not generally come from the philosophy underlying the precise methods that are used. Instead, they are more likely to be due to inadequate design, biased data, inappropriate assumptions and, perhaps most important, poor scientific practice. — p.338</em></blockquote><p>And indeed the last two chapters are dedicated to ‘How things go wrong’ and ‘How we can do statistics better’. There is a lot of discussion generally right now about the replication crisis, scientific malpractice and media misrepresentation of research and these closing sections double down on the usefulness of statistics and analyse how the different systems involved in statistical research could be improved, including a practical list of ten questions to ask when confronted by a claim backed by statistical evidence.</p><p>Finally, a delicious little footnote about research findings that are communicated in the popular press:</p><blockquote><em>I sometimes follow what could be called the ‘Groucho principle’, after Groucho Marx’s paradoxical claim that he would never join a club that would have him as a member. Because the stories have gone through so many filters that encourage distortion and selection, the very fact that I am hearing a claim based on statistics is reason to disbelieve it. — p.358</em></blockquote><h3>Affiliate link</h3><p>If you purchase the book from <a href="https://bookshop.org/a/122361/9781541675704">here</a>, I will earn a small commission (thank you!)</p><h3>Further reading</h3><ul><li>A better place to start for non-practitioners with a general interest in statistics might be <a href="https://bookshop.org/a/122361/9780393347777">Naked Statistics</a>. It’s an entertaining read with great stories.</li><li>A better entry point for actually starting to learn how to apply statistics would be <a href="https://www.amazon.com.au/Statistics-without-Tears-Introduction-Non-Mathematicians/dp/0141987499/ref=sr_1_1?crid=MZ80KZOC7O3V&amp;keywords=statistics+without+tears&amp;qid=1659230396&amp;s=digital-text&amp;sprefix=statistics+without+tears%2Cdigital-text%2C378&amp;sr=1-1-catcorr">Statistics Without Tears</a>. Or pick a course online.</li><li>There is a lot of discussion right now on scientific malpractice and miscommunication. For more on this subject, try <a href="https://www.callingbullshit.org/">Calling Bullshit</a>, a publicly available lecture series which was recently turned into a book. The first lecture starts with: ‘There is SO. MUCH. BULLSHIT. So much bullshit. So much bullshit everywhere. We’re drowning in it.’</li><li>The author has made all the data from Art of Statistics with accompanying R code available in <a href="https://github.com/dspiegel29/ArtofStatistics">this public repository</a>.</li></ul><img src="https://medium.com/_/stat?event=post.clientViewed&referrerSource=full_rss&postId=ed366f4102ea" width="1" height="1" alt="">]]></content:encoded>
        </item>
        <item>
            <title><![CDATA[How to Keep Your Head While Losing Money]]></title>
            <link>https://medium.com/swlh/how-to-keep-your-head-while-losing-money-b02e0a771f89?source=rss-8cd5173a44bc------2</link>
            <guid isPermaLink="false">https://medium.com/p/b02e0a771f89</guid>
            <category><![CDATA[growth]]></category>
            <category><![CDATA[startup]]></category>
            <category><![CDATA[business]]></category>
            <category><![CDATA[decision-making]]></category>
            <category><![CDATA[startup-lessons]]></category>
            <dc:creator><![CDATA[Junta Sekimori]]></dc:creator>
            <pubDate>Wed, 21 Jul 2021 07:47:23 GMT</pubDate>
            <atom:updated>2026-04-15T00:07:54.900Z</atom:updated>
            <content:encoded><![CDATA[<h3>Good decisions over outcomes: how to keep your head when things don’t work out</h3><h4>A bad business outcome can feel terrifying but it does not necessarily imply a bad decision. Be rational and be strong.</h4><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/0*PvWqfT6wpP89riO1" /><figcaption>Photo by <a href="https://unsplash.com/@anakin1814?utm_source=medium&amp;utm_medium=referral">Gary Meulemans</a> on <a href="https://unsplash.com/?utm_source=medium&amp;utm_medium=referral">Unsplash</a></figcaption></figure><p>It was a dreamy summer day outside as I retched into a toilet bowl in a state of panic about $3000 I had just lost for a startup I was helping grow.</p><p>I had moved to Sydney a week earlier on an impulse, acting quickly on an agreement with my partner that if ever an opportunity arose to work and live abroad for a stint, we’d drop everything and take it. The time for ‘settling down’ was approaching and if not now, we may never get the chance.</p><p>It was the beginning of the Australian summer and business was ramping up towards the Christmas peak season for the London startup. I was jet-lagged, everyone was in shorts and ‘thongs’ in November, and everything was upside down as I woke from a lunchtime nap to discover Facebook had gobbled up the largest daily budget I had ever committed in a matter of hours with little to show for it.</p><p>I am naturally risk-averse and had been carefully controlling my spend, making regular adjustments to keep the performance within an acceptable CPA ceiling. It was the early days of Facebook news feed ads when large brands still hadn’t made up their mind about them, leaving small businesses with big opportunities to acquire customers on the cheap. The ease with which you could organise the media spend targeting any of Facebook’s <a href="https://www.statista.com/statistics/264810/number-of-monthly-active-facebook-users-worldwide/">billion plus users</a> worldwide meant that anyone with an enterprising spirit could do it from anywhere in the world with an internet connection. It was a golden age of Facebook advertising for nimble startups and the dawn of the digital nomad.</p><p>And so here I was on the other side of the world, having a nice time in my shorts and thongs in a coworking space, gambling away money entrusted to me by my team back in our cold East London warehouse, back in the real world.</p><p>The startup was <em>Lost My Name,</em> whose offering was a single personalisable children’s book. How could a book be a business, I had thought, a couple of years earlier when I first met the founders, their collateral at that stage little more than <a href="https://web.archive.org/web/20121031013848/http://www.lostmy.name/">a landing page</a> to register interest. And yet in that time, they secured funding first on Dragon’s Den and most recently some investment from Google Ventures. They were approaching <a href="https://www.standard.co.uk/tech/dragons-den-recordbreakers-lost-my-name-sell-millionth-book-a3128936.html">the sale of their millionth book</a> and I had changed my mind about what a business could be.</p><h3>It felt bad</h3><p>Back at my hotdesk, I turned off all media activity and wrote Asi a carefully worded email with cold fingers. It would be a few hours until he woke up for the day. I felt I was done for, this sunny life of a modern digital professional now a stupid fiction, the others tapping away at their Macbook Airs in the coworking space no longer my kin and suddenly foreign-seeming. I had betrayed everyone’s trust and felt sick.</p><p>I explained the decisions I had made and my learnings, trying to rationalise an outcome that I couldn’t understand at the time but could later see is part of the volatility of any algorithmic trading, part of the job. I clearly wasn’t emotionally ready for making these kinds of decisions at the time but Asi was and later sent back an email, worded kindly, reassuring me that he didn’t fault my decision-making despite the outcome and that these ups and downs were part of running a business. We can’t control everything but we can control our reactions — don’t worry. I could eat again.</p><p>A month or so later, in the last week before the pre-Christmas delivery deadline, I ordered media spends well above $100,000 every day, culminating in a total spend across the festive period of over $1m, just on Facebook, and often in bare feet. I learned to hold my nerve and to trust my decision making, or rather to see that my work is to make the best possible decisions given the circumstances rather than to produce good outcomes per se, which are not totally in my control. In the long run, good decision-making will pay off and we need to find ways to keep going in the meantime, whether in human spirit or startup funding.</p><h3>Decision-making is a science, but owning a decision takes courage</h3><p>Lost My Name has since rebranded to <a href="https://www.wonderbly.com/">Wonderbly</a>, their books top the UK charts and <a href="https://www.thebookseller.com/news/wonderbly-well-equipped-china-bow-after-striking-deal-ppmg-1070536">sell in over 200 countries</a>. I am still in Sydney six years on with a second child now on the way and have focused my work in data analytics, currently managing a team of analysts at a supermarket.</p><p>I am drawn to the idea of data analytics being a discipline of ‘decision intelligence’, as the wonderful Cassie Kozyrkov from Google has been pioneering, and consider it the objective of my work to guide decision-makers away from biases and towards evidence (data) that they can use to make informed decisions. Practising this is to cultivate expertise in not just the technical handling of data but also the human biases, and the lessons are useful for life outside of business too.</p><p>In my work for Lost My Name, I lost my composure over a bad outcome and, had I been alone in deciding how to proceed next, I might have timidly wound down the advertising while I regained my confidence. Asi’s backing meant that I got back on my feet and continued to trust the logic of my decision-making.</p><p>Making good decisions in the face of uncertainty is a science that we can learn and improve at (see recommended reading below). But owning the decisions can nonetheless be tough when the stakes are high and we often need to find the courage to follow through from somewhere, whether from within ourselves or with the support of others like Asi.</p><p>Key to achieving this courage is to train our minds to continually focus on decisions and not outcomes, recognising that a bad outcome does not necessarily imply we should have made a different decision. We need to stay resolute in the face of the temptation to feel bad about an outcome and take the wrong lessons from it.</p><p>Avoid outcome bias. To repeat <a href="https://towardsdatascience.com/focus-on-decisions-not-outcomes-bf6e99cf5e4f">Cassie’s example</a>, would you rather gamble $100 on correctly guessing a coin flip or the outcome of a fair 6-sided die? The right bet is the coin flip which has a one in two outcome as opposed to one in six. But what if the coin lands the other way and the die hits the number you had in mind — should you have made the other decision?</p><blockquote>No. No, you shouldn’t have. Definitely not. That’s outcome bias and it’s a habit you should kick as soon as you can — especially if you have any hopes of being a good leader or decision-maker. [<a href="https://towardsdatascience.com/focus-on-decisions-not-outcomes-bf6e99cf5e4f">Cassie Kozyrkov</a>]</blockquote><p>And obvious though this might seem in this simple example, it might not be so obvious when the decision-making criteria are more complex, like in the case of managing advertising spend in a dynamic market. So amidst all this complexity, how can you be sure your decision is the rational one and that you haven’t missed anything?</p><h3>Building confidence in a decision</h3><p>The thing about decisions is that you wouldn’t need to make them if the outcomes were certain. So by definition, there are no perfect decisions and while you could endlessly turn over the data to inform your decisions, you are always bound by constraints in time, resources and uncertain facts.</p><p>For any decision, you have to make, try to find flaws in your arguments. Imagine you’re having to explain your thinking to a harsh critic — what counters would they throw back at you? It might help you to think of <a href="https://www.youtube.com/watch?v=vNDYUlxNIAA">the Socratic Method</a> when critiquing any argument, including your own, to put your various assumptions to the test. You might find that something you took for granted actually might not be so valid when you actively look for ways to dismantle it.</p><p>Ideally, you will also have the opportunity to pass your thinking by someone else to help find any blind spots. If so, make sure that they feel comfortable critiquing your thinking and that you are actively seeking to change your mind based on better arguments. It’s not you they are attacking, but your arguments, and that is a gift. Once nobody you have access to can propose a better alternative in the given timeframe, you are good to go.</p><p>This is a mindset particularly important to early startup growth because all you have are hypotheses. Your product or business is not established enough to have a playbook you can just follow, so you need to think critically about how to prioritise bets from the myriad possibilities.</p><p>And as a founder, be sure to foster a culture that celebrates good decisions and not just good outcomes. Stay strong for your team when they encounter failure and prop them up with emotional support, even if they’re wearing thongs on the wrong side of the world.</p><h3>Further reading for improving your decision intelligence</h3><p>I can’t recommend Farnam Street highly enough for their stimulating discussions on human biases, practical mental models and how to improve your decision-making abilities. A good place to start is their page on <a href="https://fs.blog/smart-decisions/">How to Make Smart Decisions Without Getting Lucky</a>.</p><p>Cassie Kozyrkov has been another great source of learning and inspiration for me as well, and though her writing targets the analytics community, much of it is accessible to a general audience. Start with her <a href="https://towardsdatascience.com/introduction-to-decision-intelligence-5d147ddab767">Introduction to Decision Intelligence</a>.</p><img src="https://medium.com/_/stat?event=post.clientViewed&referrerSource=full_rss&postId=b02e0a771f89" width="1" height="1" alt=""><hr><p><a href="https://medium.com/swlh/how-to-keep-your-head-while-losing-money-b02e0a771f89">How to Keep Your Head While Losing Money</a> was originally published in <a href="https://medium.com/swlh">The Startup</a> on Medium, where people are continuing the conversation by highlighting and responding to this story.</p>]]></content:encoded>
        </item>
        <item>
            <title><![CDATA[How to start a career as a data analyst from a background in the arts, and why]]></title>
            <link>https://medium.com/data-science/how-to-start-a-career-as-a-data-analyst-from-a-background-in-the-arts-and-why-913b3ac6710d?source=rss-8cd5173a44bc------2</link>
            <guid isPermaLink="false">https://medium.com/p/913b3ac6710d</guid>
            <category><![CDATA[data-analysis]]></category>
            <category><![CDATA[careers]]></category>
            <category><![CDATA[career-advice]]></category>
            <category><![CDATA[stem]]></category>
            <category><![CDATA[education]]></category>
            <dc:creator><![CDATA[Junta Sekimori]]></dc:creator>
            <pubDate>Sun, 11 Jul 2021 22:02:56 GMT</pubDate>
            <atom:updated>2021-07-12T12:02:58.740Z</atom:updated>
            <content:encoded><![CDATA[<h3>How to Start a Career as a Data Analyst from the Arts, and Why</h3><h4>What you have that STEM graduates might not and what ‘big data’ has to gain from the arts</h4><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/0*Jzrara7FCtIGJUP6" /><figcaption>Photo by <a href="https://unsplash.com/@fourcolourblack?utm_source=medium&amp;utm_medium=referral">Adam Wilson</a> on <a href="https://unsplash.com/?utm_source=medium&amp;utm_medium=referral">Unsplash</a></figcaption></figure><p><em>Isn’t it annoying when engineering students call themselves engineers? They’re getting ahead of themselves. It’s not like you hear medical students calling themselves doctors or arts students calling themselves baristas!</em></p><p>So goes one of the jokes about arts degrees (<a href="https://www.reddit.com/r/Jokes/comments/5jtaw4/isnt_it_annoying_when_engineering_students_call/">source</a>), the easy path through education for the lazy or the thick that’s a lot of fun and leads nowhere.</p><p>The place where it led for me, a graduate in French, was a career in analytics which I have found to be rewarding and well-suited to the abilities I gained during my time at university. It’s not an obvious connection to many but one I want to encourage more of in the world.</p><p>I have been in the field for over a decade now and have experienced what it is to develop professional skills in the field from a starting point of no specific training and what it is to practise analytics across agencies, start-ups and large corporates in London and Sydney.</p><p>For the last three years, I have been in a position to recruit analysts for my corporate employers and have been able to challenge the norm for candidates to come with a degree in a STEM subject, with evidence from my own experience that analytics in business has ample use for aptitudes gained in the study of the humanities (I prefer the term ‘humanities’ to ‘arts’ but I am using them interchangeably here).</p><p>In this article, I will attempt to explain what these aptitudes are, how they fit into analytics and where to start for those considering a career in analytics from non-STEM backgrounds. I hope to inspire graduates with the right inclinations to give it a shot, convince hiring managers to think more openly about the profiles they consider, and play a part generally in bringing the humanities and the sciences closer together.</p><h3>What is a ‘data analyst’?</h3><p>Firstly, what do I mean by a data analyst here? What does one do with their time?</p><p>I would say it’s fair to characterise the world of analytics as consisting of three core disciplines right now, typically split into the labels of ‘Data Analyst’, ‘Data Engineer’ and ‘Data Scientist’. In the simplest terms, the Data Analyst explores data for insights that can be used for business decision making, the Data Engineer manages the computing tools necessary for the collection, storage and access of ‘big’ data, and the Data Scientist builds software that can be used by the business to make predictions, personalise recommendations and more.</p><p>I won’t describe the other two disciplines in any more detail here, suffice to say that they are specialised and deeply technical roles that would probably not suit a generalist graduate (though the lines are often blurred at an entry level). But for more on the distinctions between the three disciplines, <a href="https://towardsdatascience.com/data-analyst-vs-data-scientist-a83af97ad472">here</a> is a good article that compares job ads for each role.</p><p>The Data Analyst role on the other hand is more accessible because <strong>novices with a comparatively simple knowledge of techniques can bring value to an organisation through other aptitudes </strong>such as good communication or an eye for insight. We can liken the role to a journalist uncovering a story, building a body of evidence to prove a hypothesis and captivating their readers through the use of persuasive storytelling. For the Data Analyst, the evidence would generally be based on numerical data and the storytelling led by data visualisation but the underlying quest and patterns of thinking are similar. (For a classic example of engaging data storytelling, see <a href="https://www.ted.com/talks/hans_rosling_the_best_stats_you_ve_ever_seen">Hans Rosling’s TED talk</a>).</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/0*iMPKK0k0IawzJYp2" /><figcaption>Photo by <a href="https://unsplash.com/@casparrubin?utm_source=medium&amp;utm_medium=referral">Caspar Camille Rubin</a> on <a href="https://unsplash.com/?utm_source=medium&amp;utm_medium=referral">Unsplash</a></figcaption></figure><p>Stories concealed in the data might be a spike in demand for a product that can be tied to certain circumstances (like food deliveries on rainy days) or the particular appeal of a product to certain types of audiences (like coffee for the cycling community). A business decision the analyst could help to make in these instances might be to increase resources for next time we anticipate such an event or to target advertising to those communities.</p><p>The tools used for data analysis typically involve spreadsheets (e.g. Excel, Google Sheets), a data visualisation tool (e.g. Tableau, Google Data Studio) and SQL code. You can find more on how to get started with these tools later on in the article. Python and R are also referenced a lot in connection to the role and are particularly useful for statistical analysis and importing publicly available data from the internet, though I would say they are inessential to begin with and a nice-to-have even at senior levels despite their prevalence in job ads.</p><p>In a typical week, an experienced data analyst might:</p><ul><li>Meet with decision makers in their organisation to discuss questions they have (e.g. <em>what made my last marketing campaign so successful? What are the reasons people don’t repeat purchases with us?</em>) and consult on approaches to answering them using available data.</li><li>Explore and clean datasets using any of the tools outlined above. ‘Cleaning’ is the process of getting messy data to line up and this is typically what takes the most time in analysing a dataset. (To give a simple example, a column of data showing customers’ gender might mix ‘F’ and ‘female’ and ‘Female’ which all indicate the same thing but are different as far as the data is concerned. To clean this, we might convert all of these to ‘F’.)</li><li>Spend time doing research to understand how the data they are working with is collected. For example, in a dataset of marketing results with revenue broken down by the channels that drove the purchases, how exactly are we attributing the purchase to a social media ad, an email newsletter etc? What if the customer saw multiple ads from multiple channels before they chose to make the purchase?</li><li>Design experiments to test hypotheses.</li><li>Present findings to decision makers with the use of visual storytelling and make suggestions on actions to take based on the insight.</li></ul><p>As an analyst deepens their technical capabilities and accumulates domain knowledge, the potential complexity of their analysis will increase, but <strong>none of this will be effective if their audience fails to understand them or glimpse the ‘so what?’ behind all the analysis.</strong> And this is where the ability to see the wood for the trees and communicate persuasively becomes paramount. It is a hurdle that trips up many of the technically brilliant types and a key area in which a humanities graduate might find they have an advantage.</p><p>What else can they bring?</p><h3>What do the humanities bring to data analytics?</h3><p>In 2020, following the Australian government’s defunding of the humanities in tertiary education, the Guardian Australia made <a href="https://www.theguardian.com/commentisfree/2020/jun/19/what-is-the-value-of-an-arts-degree-tell-us-where-yours-took-you">a shout out for comments</a> to humanities graduates asking how their degree equipped them for work.</p><p>What do you think the core themes were in the comments? Can you guess which two-word term encapsulating the values of the humanities came up more than any other across all the comments (90 times across 484 comments to be exact)?</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/0*CrLtrGsAgfiBj2S8" /><figcaption>Photo by <a href="https://unsplash.com/@giamboscaro?utm_source=medium&amp;utm_medium=referral">Giammarco</a> on <a href="https://unsplash.com/?utm_source=medium&amp;utm_medium=referral">Unsplash</a></figcaption></figure><p>Before we get to the answer, here is a comment that exemplifies the main themes that came through from readers:</p><blockquote>The degree gave me the tools to think for myself, question everything that I was being told and come to my own conclusions about what was really going on[…] Communication skills, being able to write clearly, articulately and persuasively, are consistently ranked as one of the most important attributes by employers looking for new employees. Strangely, however, they don’t seem to make the connection between the skills they’re looking for and arts graduates. [Bizkit]</blockquote><p>And here is one from a commenter working in a numerate field we tend to associate with STEM backgrounds:</p><blockquote>I majored in English and Classical literature. I spent most of my career in funds management. I ran two for major listed and private companies and spent over a decade with one of Australia’s leading fund managers. My lack of accounting training was more than compensated for by my critical assessment skills. Arts degrees teach you how to learn and think critically — more people should try it.<em> </em>[mganci23]</blockquote><p>And there it is: ‘think critically’. In a world where we are faced with a deluge of information and uncertain facts, how do we know what to trust and how to make decisions?</p><p>The term and its variations came up 90 times across 484 top-level comments, with much of the discussion in the other comments also drawing on the same idea. It’s the top bigram (two-word combination) by a long shot, excluding the terms ‘arts degree’ and ‘humanities degree’, and in a count of all single word frequencies (which was less useful here for picking out key themes), variations on ‘thinking’ ranked notably high. Here is the <a href="https://docs.google.com/spreadsheets/d/1C36NTWBkoVGYi33hPYFbgrEzr-E4qRi6h3jBwnQ0G10">dataset</a> if you would like to take a look.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/0*Q-M7jbN_F0a2tWzu" /><figcaption>Image by author</figcaption></figure><p>So why is ‘critical thinking’ a hallmark of the humanities, and is it really not acquired as part of a STEM education?</p><p>I imagine there will be STEM graduates ready to staunchly defend their critical thinking abilities gained from their time studying or otherwise, but I think there is enough evidence to draw on that suggests there is some shortage of it in the way STEM is broadly taught. There is a much discussed <a href="https://www.vox.com/future-perfect/21504366/science-replication-crisis-peer-review-statistics">replication crisis</a> unfolding in the social sciences at the moment which casts doubt on the fundamental approach that scientists are taking in drawing conclusions, and it seems like parts of the scientific community have become habituated to using statistical tools <a href="https://fivethirtyeight.com/features/not-even-scientists-can-easily-explain-p-values/">they don’t fully understand</a>. David Epstein, <a href="https://www.ted.com/speakers/david_epstein">a critic of poor scientific research</a>, suggests that shallow thinking is endemic in science from the way it is learned in universities, claiming in his book <em>Range</em>:</p><blockquote>Like many a grad student, I had a big database and hit a computer button to run a common statistical analysis, never having been taught to think deeply (or at all) about how that statistical analysis even worked. The stat program spit out a number summarily deemed ‘statistically significant.’ Unfortunately, it was almost certainly a false positive because I did not understand the limitations of the statistical test in the context in which I applied it. Nor did the scientists who reviewed the work[…] I rushed into extremely specialized scientific research without having learned scientific reasoning. (And then I was rewarded for it, with a master’s degree).</blockquote><p>And it is easy to imagine in degrees like engineering and computer science that students are directed towards rules-based thinking, building up knowledge of how things fit together and dexterity in solving logical problems — valuable skills, no doubt. <strong>But how far will these skills help them in the world of human decision making where there often aren’t straight answers to problems, or there are missing facts, or there are severe limits to time in which to come to conclusions?</strong></p><p>How to weigh up evidence amidst uncertainty, form a position and explain it is the domain of the humanities and as the tide of STEM has been rising, so too is emerging a counter pressure from the humanities urging for balance. A good focal point for this can be found in the <a href="https://theconversation.com/steam-not-stem-why-scientists-need-arts-training-89788">STEAM movement</a> which advocates for the adding of ‘A’ for ‘Arts’ to STEM education to cultivate more human-centric thinking.</p><p>But how does this relate to data analytics? Isn’t the use of data basically a maths problem? No, and this is where there is still a massive misperception in what analytics entails in practice. <strong>The biggest misperception, which the industry benefits from upholding, is that what the data tells us are objective facts. </strong>In reality, there is uncertainty and subjectivity in the use of data (see <a href="https://towardsdatascience.com/the-perception-of-objectivity-in-data-1abacf195b29">Jonathan Choi’s discussion</a> for a good starting point on this), with plenty of room for bending data to artificially support an agenda as highlighted in books such as <a href="https://www.goodreads.com/book/show/28504537-a-field-guide-to-lies"><em>A Field Guide to Lies</em></a><em> </em>(also published under the title <em>Weaponized Lies</em>) and <a href="https://www.goodreads.com/book/show/48889983-calling-bullshit"><em>Calling Bullshit</em>.</a> Experienced analysts in big organisations will likely be all too familiar with the request to back up a statement with supporting data, with the opportunity perhaps to challenge it but not enough time available to do so.</p><p>Often when talking about the role the humanities might play in data analysis, we jump to the communication aspects, but another key aptitude is a mindset of questioning how the data will be used (leading to the ‘so what?’) and, just as importantly, where the data has come from. There is a dictum in computer science <em>Garbage In Garbage Out</em>, but rare is the analyst who questions how valid the data is in the first place or makes the data’s origin a subject of their inquiry. There are often biases or problems in the way data is collected in the first place that might change our interpretation of the data and, obvious though this might sound, analysts all too often fail to critically examine their source of information.</p><p><strong>Critical thinking is a habit of mind that is difficult to teach. The techniques for manipulating data, whether through code or other tools, are easier.</strong></p><h3>Why job ads for data analyst positions ask for a STEM degree</h3><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/0*3NoqdVqXFJAL95DH" /><figcaption>Photo by <a href="https://unsplash.com/@thisisengineering?utm_source=medium&amp;utm_medium=referral">ThisisEngineering RAEng</a> on <a href="https://unsplash.com/?utm_source=medium&amp;utm_medium=referral">Unsplash</a></figcaption></figure><p>Every job in analytics I have been offered since my first has stipulated the requirement for a degree in maths, computer science, statistics, economics and the like, which I don’t have. But this has never got in the way.</p><p>In Australia, where the ethos of going to university seems to be more about vocational training than it is in the UK, colleagues have been surprised about my degree background. When I make new connections on LinkedIn, I am curious to check their educational backgrounds, and I can’t think of a single analyst I’ve met in Australia who doesn’t have a STEM or business-related degree. In the UK, where employers are less prescriptive over undergraduate fields of study, there is more of a mix of backgrounds. (Curiously, software developers seem to come from far more eccentric backgrounds).</p><p>I don’t doubt that STEM degrees provide some practical tools for the job and a leg up for securing the first role, but I want others to know that most of these degrees don’t directly equip graduates for the job. Mathematicians are not taught how to evaluate the performance of marketing campaigns. Computer scientists are not taught how to conduct A/B tests in a business. They pick these things up on the job and <strong>employers look for graduates who are quick to learn, pleasant to teach and who will get on well with everyone as they grow into the role.</strong></p><p>The technical skills essential for data analysis can be learned on the job to a sufficiently useful standard fairly quickly and career progression very soon starts to focus on building domain knowledge (i.e. how your industry works, what the latest trends are) and improving ‘soft skills’, which university degrees do not cover in any relevant depth. That’s not to say that STEM is irrelevant to this role — the scientific method and certain specifics such as an understanding of statistics are very much useful — but my point is that it doesn’t directly prepare you for this job and there are qualities from the humanities that are also useful and undervalued. Ultimately, <strong>it is a role that suits people who have range</strong>.</p><p>And it might help job seekers to understand that the primary reason STEM is demanded in job ads is they act as a filter that helps reduce the number of applications that recruiters have to review. For the reason that anyone is free to apply for a publicly advertised job, big companies might receive hundreds of applications and it would simply be impossible to review every one, so they need a way to maximise their chance of finding good candidates within their time limits.</p><p>Adding the requirement of a STEM degree up front for this purpose makes the assumption that there is a higher likelihood of a STEM applicant being a good fit for the role than a humanities applicant. The assumed composition of the applicant pool might look something like the below. (The distributions are hypothetical and the aim of this is to help conceptualise the decision making heuristics that hiring managers use. It is not intended as insight into the sizes and shapes of the talent pool.)</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/673/0*GXrq6GfcrGDPrv_K" /><figcaption>Image by author</figcaption></figure><p>The bars represent distributions of humanities and STEM graduates in terms of their likelihood of success in the role of a junior data analyst.</p><p>We are assuming the average STEM grad has a higher likelihood of success in the role than the average humanities grad, but the suitability of graduates will be variable within each group and some humanities grads will be more suitable than many STEM grads.</p><p>Despite some good humanities candidates at the upper end of the distribution, if employers are limited in the time they can take to consider applicants, picking only from the green pool would be a reasonable time-saving measure.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/646/0*3aJH1lg4SqmORkLx" /><figcaption>Image by author</figcaption></figure><p>But what about the exceptional humanities graduates in the right tail of the yellow distribution? Unfortunately, missing out on them is the tradeoff of using the filter. <strong>They are rejected not because they are actually unsuitable for the role but because they are unsuitable for <em>the employer’s consideration</em> given their time constraints and the heuristics they’ve put in play</strong>. (As an interesting aside, <a href="https://www.businessinsider.com.au/why-nigerian-scam-emails-are-obvious-2014-5">it has been suggested</a> a similar process is action with the bad English grammar of email scammers — those who aren’t likely to fall for it will be suspicious of the poor writing and drop off instantly, thus minimising the time the scammers might waste in follow-up work with bait that ultimately doesn’t pay.)</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/682/0*9kMHW2dcFDxkOMeR" /><figcaption>Image by author</figcaption></figure><p>So, job seekers, understand that a failure to get past the screening stage of an application may not be about your potential in the role and career but about the practical limitations of employers considering every application. Sadly, the current heuristic for using STEM is an effective one so those who fall victim to it need another route in.</p><h3>Getting a first job in analytics</h3><p>Regardless of your background, here are some tips for entering the field if the ‘front door’ is proving difficult.</p><h4>1. Start studying online</h4><p>My entry into analytics was aided by some self-teaching via books. The demand for data analysts in industry has since ballooned and so has the amount of excellent learning resources online. Diving into some learning will not only give you skills that will be directly useful in a data analytics career but also test whether you will indeed enjoy the career or not.</p><p>My go-to free online resource I point beginners to is <a href="https://mode.com/sql-tutorial/introduction-to-sql/">Mode’s excellent primer on SQL</a>. For more expansive, paid options I would suggest diving into <a href="https://www.datacamp.com/courses/introduction-to-sql">Datacamp’s</a> interactive SQL tutorials or <a href="https://grow.google/dataanalytics">Google’s data analytics course on Coursera</a> which seems to be <a href="https://www.youtube.com/results?search_query=google+data+analytics+certificate">getting a lot of attention</a> right now.</p><p>Give yourself about 10 hours of study a week for 3 months and you will have got a good taste of what is involved and some useful skills to exhibit in your job applications, as well as evidence of your motivation.</p><h4>2. Network</h4><p>Follow analytics agencies and individuals working in analytics on LinkedIn and Twitter. Don’t hesitate to ask strangers for advice and don’t be discouraged if you mostly don’t hear back. What they post will give you a good idea of the topics that are currently front of mind in the community and there will often be links to free webinars and local events you can join. Look up analytics events in your local city on meetup.com (I recommend ‘Digital Analytics Wednesday’ or ‘Data &amp; Analytics Wednesday’ which takes place in many major cities around the world once a month). Ask people you know to introduce you to analysts that they know and get their advice.</p><h4>3. Enter analytics via other roles</h4><p>You would be hard pressed to find a business function in which data isn’t useful. Don’t feel like you’re taking a wrong turn by applying for other roles and, once in, find ways to get exposure to data. Volunteer to clean up a spreadsheet or challenge yourself to improve a part of your function with the use of data. Perhaps you could organise an internal survey and create a pack for your team visualising the results. Make contact with people in your organisation who are performing data analysis in their functions. Keep at the online courses in the evenings.</p><h3>Closing thoughts</h3><p>I intended this article primarily for job seekers coming from the arts and the fact that you found your way to this article probably means that you have already considered this as a direction in your life. I hope I have been able to inspire you to take the plunge and convince you that this could be a viable option for you despite all the signs pointing otherwise.</p><p>For others, particularly in business, I hope this has given you some food for thought on the challenges of data analysis in a dynamic, fast-moving environment and the need for more critical thinking.</p><img src="https://medium.com/_/stat?event=post.clientViewed&referrerSource=full_rss&postId=913b3ac6710d" width="1" height="1" alt=""><hr><p><a href="https://medium.com/data-science/how-to-start-a-career-as-a-data-analyst-from-a-background-in-the-arts-and-why-913b3ac6710d">How to start a career as a data analyst from a background in the arts, and why</a> was originally published in <a href="https://medium.com/data-science">TDS Archive</a> on Medium, where people are continuing the conversation by highlighting and responding to this story.</p>]]></content:encoded>
        </item>
    </channel>
</rss>