<?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 Joe Moran on Medium]]></title>
        <description><![CDATA[Stories by Joe Moran on Medium]]></description>
        <link>https://medium.com/@joe-moran?source=rss-1d1feac897f6------2</link>
        <image>
            <url>https://cdn-images-1.medium.com/fit/c/150/150/0*WU2kKqhTmq3OlClN.jpg</url>
            <title>Stories by Joe Moran on Medium</title>
            <link>https://medium.com/@joe-moran?source=rss-1d1feac897f6------2</link>
        </image>
        <generator>Medium</generator>
        <lastBuildDate>Tue, 17 Nov 2026 13:22:54 GMT</lastBuildDate>
        <atom:link href="https://medium.com/@joe-moran/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[How do we transform enemies into friends?]]></title>
            <link>https://joe-moran.medium.com/how-do-we-transform-enemies-into-friends-11c7b7b2d9e6?source=rss-1d1feac897f6------2</link>
            <guid isPermaLink="false">https://medium.com/p/11c7b7b2d9e6</guid>
            <category><![CDATA[usa]]></category>
            <category><![CDATA[democracy]]></category>
            <category><![CDATA[empathy]]></category>
            <category><![CDATA[social-media]]></category>
            <category><![CDATA[unrest]]></category>
            <dc:creator><![CDATA[Joe Moran]]></dc:creator>
            <pubDate>Tue, 19 Jan 2021 16:44:54 GMT</pubDate>
            <atom:updated>2021-01-19T16:44:54.403Z</atom:updated>
            <content:encoded><![CDATA[<p>I’ve been logged out of social media for 21 days now. It was a highly consequential 21 days, with insurrectionists attempting to overthrow the US government by storming the Capitol, egged on by Trump who later had his Twitter account deactivated and got impeached for it. I will leave you to decide which is a worse outcome for him. Free speech enthusiasts got their knickers in a twist about it, and free people enthusiasts got rightly angry about both <em>that </em>and the Capitol insurrection. It is certainly chilling that Twitter et al. have the power to silence people. The actions by them and others last week removed the band aid that prevented us from seeing how bad our systems of speech and checks/balances have become. While I don’t want Twitter in charge of who gets a platform and who doesn’t, I do want there to be consequences for being able to spew misinformation, foment insurrection, and generally be a fountain of lies. Media and science mistrust is at an all-time high. People I thought were smart have become radicalised and now believe nothing that “legacy” media say, and we’ve come to a very dark place where facts seemingly have ceased to exist. I don’t think this is OK.</p><p>While all of that has been a LOT, I’m feeling now that logging out was the best decision I made. I’ve gotten my news from the news: Washington Post, CNN, BBC, NPR. And yes, there’s a liberal slant in some of those places, but they remain, news. WaPo publishes opinion by conservatives regularly, that I read. BBC and NPR attempt to both-sides things in quite reasonable terms. CNN, for all its faults, has reporters who report on real things actually happening. Getting my news this way releases me from the outrage machine. On Twitter, I mostly followed liberal folks, as I too am left-leaning. But the constant torrent of invective is quite twisting. I saw every news headline through the progressive scope of one or more different factions of the left; people who filter the facts through their lens, refracting reality so that light reaches only their anger-generating places.</p><p>Interestingly, I’ve still been exposed to social media by various friends, family and contacts. Hearing what is going on at a second-hand remove has been quite curious — I don’t get the same anger rising, the emotion threatening to take over my typing fingers. Instead, I start to reason more, empathise, and consider why and how the crazy conservative on the other end of that comment might have arrived at their erroneous conclusion.</p><p>And things become a bit more clear. Here is what I’ve heard; I’m playing it back to see if it’s a true representation, not to endorse it.</p><ul><li>Conservatives have felt for some time that the public sphere is being “controlled” by a certain kind of liberal orthodoxy that allows only for a certain kind of speech. One must make statements that support marginalised folks, or be silenced. Marginalised folks include the LGBTQ community, non-white minorities, women, displaced peoples, less-abled folks, and any and all intersections therein. If you make a statement that impinges upon (or appears to) the rights of any of these folks, you will be canceled. Cancelation means silencing and ostracism through a public outcry of shame. Therefore, conservatives feel they are not allowed to participate in Public Life in the ways that they have previously.</li><li>Donald Trump represented a salve to this; he got away with saying politically incorrect things — “telling it like it is” — and so attained a hero-like status to those who felt like they could no longer express themselves. Any attempt to silence Donald Trump is a direct attempt to silence <em>them</em>.</li></ul><p>This thoroughly modern disapproval apparatus, and the reactions it has led to, have been coupled to a long-seething, slow moving catastrophe of the middle class. Beginning with Reagan’s Neoliberal economic trickle down ideas of the 1980s, the Republican Party began to support policies that redistributed money upwards; low capital gains taxes, free trade, offshoring, deregulation, etc. The Democrats did not oppose this, instead they went right along, becoming a pro-business party indistinguishable from the GOP in their support of policies that benefit big business while screwing the worker. Factory closings, restructuring, lay-offs, a gutting of manufacturing. These things all occurred on both Republicans’ and Democrats’ watch since 1980 right up until now.</p><p>And the conversation at the national Democratic level for at least 5–6 years has been about progressivism — how do we make reparations for folks left behind? How do we provide basic protections for people whose very existence has been illegal? These are good and necessary fights. But the Democratic Party has ignored the working person. The GOP has not. In fact, the GOP has gotten very good at telling the working person that the reason they have fallen behind is precisely because of these marginalised people — people of colour and immigrants, who squeeze out working white folks’ livelihoods and culture. It’s been a huge lie, but an effective lie, nonetheless.</p><p>Meanwhile, Democrats have dug in their heels. Instead of attacking income inequality and facilitating dignity of the worker, they have responded to charges of naked progressivism by engaging in naked progressivism. Again, getting marginalised folks on an equal footing; creating conditions that allow for equality of opportunity is a critical goal. But it won’t be achieved by pandering to business and ignoring the plight of the worker. In fact, people in many different communities are “the worker”. Black and brown and white folks alike have seen their spending power diminish and the American dream evaporate over the last 40 years. The Republicans have made hay by conjuring policies that created these conditions in their left hand, while pulling an immigrant rabbit out of a hat with their right. The Democrats have indirectly told working folks that they get nothing by trying to out-Republican the GOP on business, and they have indirectly told <em>white</em> working folks that they don’t matter by largely pursuing progressive policies designed to equalise opportunity.</p><p>But what if the Democrats realised that folks from all cultures are hurt by Neoliberalism? What if they passed a $15 minimum wage, equalised capital gains and income taxes, and put people to work with a new deal based on moving to a green industry and economy? Prosperity of the working and middle classes would result. And in that prosperity begins the budding realisation that our fellow Americans are not the enemy.</p><figure><img alt="Hands reaching for one another" src="https://cdn-images-1.medium.com/max/1024/1*Rn38QONacAbomjnd0-pmsA.jpeg" /><figcaption><a href="https://unsplash.com/photos/iJ2IG8ckCpA?utm_source=unsplash&amp;utm_medium=referral&amp;utm_content=creditShareLink">https://unsplash.com/photos/iJ2IG8ckCpA?utm_source=unsplash&amp;utm_medium=referral&amp;utm_content=creditShareLink</a></figcaption></figure><p>Social media has fomented the revolution of antipathy. Empathy lies bleeding while thumbs connected to the rage centres of the brain fire off uncured anger. It’s easy to lash out when you don’t have to see a person’s emotional reaction. It’s easy to dehumanize. Posting on social media, compared to debating in person, is like firing a gun instead of using a knife. Guns kill people, because people won’t get close enough to stab someone, it’s too emotional, too personal. Social media kills democracy because (most) people have too much humanity to scream obscenities directly in the face of people they disagree with.</p><p>This experiment in democratising communication has revealed much. But one thing it hasn’t done is “shown us who we really are”. I argue that it has simply divorced the ego from the id; people free to act on their basest impulses do not debate. People who share dinner and coffee debate. We have 10 or so years of polarization and extremity not because we are so different, but because the situations we have created for ourselves reward extremity and discourage understanding. It’s time to get back into the public square where we can see how our words land. Maybe we will then remember that the other is a person.</p><p>As Dr. King said: “Love is the only force capable of transforming an enemy into a friend”.</p><img src="https://medium.com/_/stat?event=post.clientViewed&referrerSource=full_rss&postId=11c7b7b2d9e6" width="1" height="1" alt="">]]></content:encoded>
        </item>
        <item>
            <title><![CDATA[Life during a pandemic]]></title>
            <link>https://joe-moran.medium.com/life-during-a-pandemic-ecf828062c7b?source=rss-1d1feac897f6------2</link>
            <guid isPermaLink="false">https://medium.com/p/ecf828062c7b</guid>
            <category><![CDATA[kindness]]></category>
            <category><![CDATA[pandemic]]></category>
            <dc:creator><![CDATA[Joe Moran]]></dc:creator>
            <pubDate>Tue, 17 Mar 2020 01:25:22 GMT</pubDate>
            <atom:updated>2020-03-17T01:33:40.291Z</atom:updated>
            <content:encoded><![CDATA[<p>The images come in waves. South Korea, Italy, Spain, France. Lockdown, shuttered stores, the army patrolling the streets. Here in the States we are told to #flattenthecurve and #canceleverything, or else we too could overload our hospitals with the sick and dying. It’s very likely still to happen.</p><p>So many of us are afraid, unsure what to tell ourselves, unsure of the near future. Is this tickle in my throat seasonal allergies, or something less…familiar? Does my slightly elevated blood pressure, or my wife’s asthma put us at high risk? What does high risk mean? How can so many carry this disease and have no symptoms, while to some this could be a death sentence? How do I get to sleep tonight?</p><p>The virus exploits our human need for connection in the most cynical of ways. The more we seek out one another, the better it spreads. The longer it takes to sicken us, the better it spreads. Humans need connection to survive — viruses figured this out a very long time ago.</p><p>Our televisions and feeds are filled with panic; panic-buying, panic-prognosticating, panic-selling. Some sit on the sidelines and ridicule the panickers. Some sit in bars and thumb their noses at the statistics, the sage advice of scientists, the very notion that curtailing the luxury of drinking a Negroni in public may save the infirm. Others seek to profit: Cyber-attacks are on the rise, capitalists stockpile sanitiser and politicians add cynical amendments to the bills meant to save us. Rumours abound. Gargling salt water will save you, drinking water will save you, spring weather will save you. Real virologists and infectious disease experts jostle on Twitter with psychopath know-it-alls who rejoice in very public displays of misinformation. Both have the same-sized platform.</p><p>Many, many hourly workers, service workers, carers, artists, musicians, performers and others face a very uncertain short to medium term. We need to come up with a plan to help people, and put stimulus where it needs to go. To the people.</p><p>The president has been an ass. He has dissembled, caring only about his reputation, or whether the tanking markets will sink his presidency come November. He has repeatedly, publicly, video-ly sought to downplay the virus, and now claims to have never said these things. He will continue to do this until he is finally, beautifully removed from office.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*dnce92tc34kv-Tt_o1v1RQ.jpeg" /></figure><p>But there’s a different story afoot too. As the rhythms of life slow, and those lucky enough to work from home stop commuting, people are learning how to be people again. Distant connections are emerging. A kind call from across the street as you walk your dog. “Hey, how are you doing?” No one fucking talks to you on the street in Boston. Italian tenors serenade their neighbourhoods. Spanish trainers give fitness classes from their roof to the apartment block opposite. Yo-Yo Ma gives an online concerto for the health-care workers on the front lines. Remote coworkers become like family, sharing smiles and jokes on the 930am virtual standup. Jigsaw puzzles and properly-sanitised board games leap from the cupboards and spring back into action. They were always ready. People are rediscovering lost projects. Hey, I wanted to learn how to paint; how about some meditation? There’s a new calm. Gluten relaxes alongside us. Notes are passed from apartment to apartment. Can I get you anything? Do you need milk? The troops are mobilising to share #COVIDkindness.</p><p>Suddenly our living rooms are shared spaces we can all be invited into, united by technology. We’ve been reminded that underneath it all, we can be kind. Let’s keep it going once the crisis is over.</p><img src="https://medium.com/_/stat?event=post.clientViewed&referrerSource=full_rss&postId=ecf828062c7b" width="1" height="1" alt="">]]></content:encoded>
        </item>
        <item>
            <title><![CDATA[Harnessing the power of purrr]]></title>
            <link>https://joe-moran.medium.com/harnessing-the-power-of-purrr-33991f2f48c3?source=rss-1d1feac897f6------2</link>
            <guid isPermaLink="false">https://medium.com/p/33991f2f48c3</guid>
            <category><![CDATA[power-analysis]]></category>
            <category><![CDATA[data-science]]></category>
            <category><![CDATA[rstats]]></category>
            <category><![CDATA[statistics]]></category>
            <dc:creator><![CDATA[Joe Moran]]></dc:creator>
            <pubDate>Fri, 08 Jun 2018 17:35:43 GMT</pubDate>
            <atom:updated>2018-06-08T17:45:41.767Z</atom:updated>
            <content:encoded><![CDATA[<p>R is a powerful language for reproducible data science work, which is particularly well-suited for statistical analysis. Recently, my company asked me to estimate the sample sizes needed to observe significant effects of a few new interventions we will soon try. This requires power analysis (which we will complete using the excellent pwr package in R). We also needed to run this analysis multiple times on similar (but different) inputs:</p><p>We wanted to test this on (a) different metrics, (hence different expected effect sizes) (b) in A/B <strong>and</strong> repeated measures (A[t2] vs. A[t1]) contexts, and (c) across a range of observable powers, so that we could develop a range of needed sample sizes.</p><p>This is clearly a job for purrr, <a href="https://medium.com/u/2f166aac6770">Hadley Wickham</a> and Lionel Henry’s excellent <a href="https://purrr.tidyverse.org/">toolkit for functional programming</a>. Within purrr, there exists a series of functions in the map family, which enable simple and elegant <strong>iteration </strong>across lists and data frames.</p><pre>library(dplyr)<br>library(purrr)<br>library(pwr)</pre><p>For this task, we need to develop a series of inputs to the pwr::pwr.t.test() function. This function takes an expected effect size, <a href="http://rpsychologist.com/d3/cohend/">Cohen’s <em>d</em></a>; a significance level (by convention <em>p</em> &lt; .05); a <em>power </em>(more on this shortly); and a <em>type</em> (one-sample, two-sample, paired).</p><p>Power refers to the likelihood of observing a significant difference between two datasets <em>if one exists</em>. Mathematically, it is 1 - ß (1 - probability Type II error). A power of 0.8 therefore means that if the hypothesized difference in means between our datasets is really there, we will observe it 80% of the time.</p><p>Let’s set up some parameters for our power analysis!</p><pre># Get some sample effect sizes (vary them around 0.2)<br>d1 &lt;- rnorm(6, 0.2, .05)</pre><pre># Perhaps we expect a stronger effect? (vary them around 0.4)<br>d2 &lt;- rnorm(6, 0.4, .05)</pre><pre># Combine them for further manipulation<br>d &lt;- c(d1, d2)</pre><pre># Range of acceptable power<br>power = c(.5, .8)</pre><pre># Different designs, &quot;two.sample&quot; = A/B test:<br>type = c(&quot;paired&quot;, &quot;two.sample&quot;)</pre><p>To get this ready for analysis, group it all up into a data frame:</p><pre># Variable code the number of levels of each factor in the test:<br>npwr &lt;- length(power); ntyp &lt;- length(type); nd &lt;- length(d)</pre><pre>df &lt;- data.frame(d = rep(d, npwr * ntyp),<br>                 power = rep(power, ntyp * nd),<br>                 type = rep(type, npwr * nd),<br>                 stringsAsFactors = FALSE)</pre><p>We want pwr.t.test to tell us what sample size n we need to achieve a particular power, so we will leave this argument NULL . Because we assume the same significance level for each test, we will set this value to 0.05 for all power calculations.</p><p>Now comes time to actually calculate the power of our different experimental designs!</p><p>purrr::pmap is a handy tool that allows you to specify your iterated function as a formula. This method allows us to explicitly call out the relevant parts of our data frame that we would like to sample from. In this example, our data frame has 3 columns:</p><ol><li>The effect sizes, <em>d</em></li><li>The acceptable power values</li><li>The kind of test (A/B or repeated measures)</li></ol><p>We’re going to use these column positions inside our formulaic call to pmap, telling R to <em>iterate </em>over these columns of the data frame to get the inputs for pwr.t.test while allowing n to remain NULL and keeping our significance level at .05 for all calculations:</p><pre># Pass our data frame and constant arguments to pmap<br># d is passed as argument 1: ..1 etc.</pre><pre>pwr_results &lt;- df %&gt;%<br>  pmap(~ pwr.t.test(n = NULL, ..1, sig.level = 0.05, ..2, ..3)</pre><p>And there we have it! Interrogating our output list, we see that we have results returned for each of the experimental scenarios we constructed:</p><pre>pwr_results[[1]]</pre><p><em>Paired t test power calculation</em></p><p><em>n = 59.76034<br> d = 0.2577413<br> sig.level = 0.05<br> power = 0.5<br> alternative = two.sided</em></p><p>Tidying up this list output is not in the scope of this post, but the functions broom:tidy, unlist, and tidyr::gather; tidyr::spread might be of use.</p><p>Thanks for reading — and as always, suggestions for code improvements are welcome! What are your tips for automating power analyses in R?</p><img src="https://medium.com/_/stat?event=post.clientViewed&referrerSource=full_rss&postId=33991f2f48c3" width="1" height="1" alt="">]]></content:encoded>
        </item>
        <item>
            <title><![CDATA[Some things I learned when I added ‘data’ to ‘scientist’.]]></title>
            <link>https://joe-moran.medium.com/some-things-i-learned-when-i-added-data-to-scientist-2f9cdae58f4?source=rss-1d1feac897f6------2</link>
            <guid isPermaLink="false">https://medium.com/p/2f9cdae58f4</guid>
            <category><![CDATA[analytics]]></category>
            <category><![CDATA[python]]></category>
            <category><![CDATA[data-science]]></category>
            <category><![CDATA[rstats]]></category>
            <dc:creator><![CDATA[Joe Moran]]></dc:creator>
            <pubDate>Fri, 16 Feb 2018 02:02:54 GMT</pubDate>
            <atom:updated>2018-02-16T12:14:46.359Z</atom:updated>
            <content:encoded><![CDATA[<p>My transition is a relatively common one these days. My background is in academic psychology; most recently I branched out into UX research with the United States Army, and over the last year I’ve been combining my UX research and data science skills at the AI startup <a href="http://www.cogitocorp.com/">Cogito</a>. My community building over this last year has seen me talking to ex-astronomers, ex-public policy advocates, and ex-economists, amongst others, but all ex-<em>quantitative scientific researchers</em>.</p><p>This post is about something I think we all share in common, which is that when you make the transition from science -&gt; data science, there’s a few things that might help you along the way. Some of these things I probably should have foreseen, and others sneak up on you, like a giraffe hiding behind a lamppost — maybe had you changed your viewing angle just a tiny bit you might have caught a glance.</p><p>While I gathered some data science skills, the path seemed clear: learn enough R and/or Python to be able to do the thing fluently and efficiently, develop enough machine learning to know what to deploy, when, and how, and make sure I know how to construct a quality visualization.</p><p>But it goes beyond that.</p><p>How do you explain to a sales exec what a scatterplot is? How do you go from the excitement of seeing a 0.4 correlation between something your company measures and some important outcome to putting that in front of the marketing team and having them say “wait, 0.4 out of 1, doesn’t that suck?”</p><p>How do you explain that the logistic model you just built predicting awesome outcome for client A might not hold up if we apply it to client B, who actually sells shoes to Merino sheep and not Clydesdale horses?</p><p>These problems all have the same root: Scientists typically speak to other scientists, who have also spent long hours at the statistical grist mill, learning the arcane ways of significance, outliers, representativeness, model selection criteria, etc. But data scientists speak to business leaders, marketers, sales people, and product managers, all of whom have a different path and their own different expertise. Below are a couple of solutions I’ve stumbled across along the way.</p><ol><li><strong>Start small.</strong> Start with the simplest analysis possible, and make the most basic case you can. Then you can add complexity as understanding grows. Here’s a bar chart showing A &gt; B. Now, here’s a boxplot showing that while those averages kind of look similar, there’s a whole lot of visitors taking longer on Site B than Site A. Now, here’s a bubbleplot showing that all those people who were taking extra time had been using our product for only five minutes.</li></ol><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/0*pGoHNEOiAUfkB8u8." /><figcaption>Start Small! (image by <a href="https://unsplash.com/@artic_studios?utm_source=medium&amp;utm_medium=referral">Daniel Hjalmarsson</a> on <a href="https://unsplash.com/?utm_source=medium&amp;utm_medium=referral">Unsplash</a>)</figcaption></figure><p><strong>2. Make no assumptions. </strong>This came out of a Twitter exchange with <a href="https://medium.com/u/56211c3483cf">Jesse Maegan</a>:</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*pK2YVPHLvsO_c-bkZ4Cacg.jpeg" /><figcaption>Mind the gap</figcaption></figure><p>This exchange reminds us that to provide actionable insights, they need to be <em>viewed as insights</em> in the mind of the consumer before they can ever have a hope of being actionable. Taking many journeys to the land of the furrowed brow has taught me to spend more time upfront querying my stakeholders. “Here’s the kind of evidence we’ll provide to answer your question. Here’s why this visualization is effective for this kind of problem, and here’s what you can expect to see if the difference you were hoping for is real.”</p><p><strong>3. Check and recheck. </strong>Once it comes time to deliver an operational insight, be sure that your internal customer understands the big picture and the small picture. Can they explain to the client what is being presented without your being in the room? Have you clarified that ‘better than’ does not necessarily mean ‘improved’, or that statistically significant does not necessarily mean ‘large’?</p><p>All of these things seem obvious in retrospect. Of course we want to make sure people can consume our work, of course we should check for understanding, and set clear expectations. But the scientific norm is that we analyse, distill, report on, and publish all our own stuff. Transitioning to data science means that many of these steps occur post-handoff, or that you need to adapt and jettison your jargon before speaking to the stakeholder.</p><p>Having learned first-hand the many ways in which you can mess up your communication as a data scientist, I hope to soon start minimising those errors. I’m under no illusions though; hard-won inside knowledge is just as hard-forgotten, and this stuff takes time.</p><p>What are your tips for making the leap from science to data science?</p><img src="https://medium.com/_/stat?event=post.clientViewed&referrerSource=full_rss&postId=2f9cdae58f4" width="1" height="1" alt="">]]></content:encoded>
        </item>
    </channel>
</rss>