<?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 Matheus Reis on Medium]]></title>
        <description><![CDATA[Stories by Matheus Reis on Medium]]></description>
        <link>https://medium.com/@matheusreis_13304?source=rss-2da90e192759------2</link>
        <image>
            <url>https://cdn-images-1.medium.com/fit/c/150/150/1*uYo_x37jEAu50t_W3AvdAw.jpeg</url>
            <title>Stories by Matheus Reis on Medium</title>
            <link>https://medium.com/@matheusreis_13304?source=rss-2da90e192759------2</link>
        </image>
        <generator>Medium</generator>
        <lastBuildDate>Fri, 20 Nov 2026 09:46:26 GMT</lastBuildDate>
        <atom:link href="https://medium.com/@matheusreis_13304/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[What happened when I tested TARGET(AI) on a real open-source software problem.]]></title>
            <link>https://medium.com/@matheusreis_13304/what-happened-when-i-tested-target-ai-on-a-real-open-source-software-problem-fb8a6ab2a763?source=rss-2da90e192759------2</link>
            <guid isPermaLink="false">https://medium.com/p/fb8a6ab2a763</guid>
            <dc:creator><![CDATA[Matheus Reis]]></dc:creator>
            <pubDate>Sun, 06 Sep 2026 16:06:06 GMT</pubDate>
            <atom:updated>2026-09-06T16:06:06.113Z</atom:updated>
            <content:encoded><![CDATA[<p>I had a plan.</p><p>It was a good plan, too.</p><p>The problem was a bug in OpenCode. Nothing dramatic. An agent defined in Markdown could lose part of its prompt configuration along the way.</p><p>I asked an AI agent to investigate it and come up with a normal engineering plan. It did.</p><p>The plan made sense. Find where the prompt was being lost. Fix the precedence. Resolve the file reference from the right place. Add tests. Done.</p><p>Then I ran the same problem through TARGET(AI).</p><p>Mostly, the two plans agreed.</p><p>That was reassuring.</p><p>Then TARGET(AI) made me look at one assumption again.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*ScBnlVw3_eLZfIXtFErmjQ.png" /><figcaption>AI generated, based on TARGET(AI)´s fix</figcaption></figure><p>The prompt could contain a relative path to another file. We had assumed that path should be resolved relative to the Markdown file that contained it.</p><p>Perfectly reasonable.</p><p>Also wrong.</p><p>When we went back to the evidence, OpenCode already had an answer to that question. Its configuration semantics pointed somewhere else: the configuration directory.</p><p>It was a small thing.</p><p>A path.</p><p>Probably a few lines of code.</p><p>But that’s exactly the kind of mistake I find interesting. The AI hadn’t failed. It had produced a sensible solution to a sensible interpretation of the problem.</p><blockquote><strong>The interpretation of an agentic AI was wrong.</strong></blockquote><p>So we changed it before the fix became the fix.</p><p>Then we tested it.</p><p>Five regression cases. Four failed before the change. All five passed afterward. The full configuration suite passed 113 tests.</p><p>The patch itself isn’t particularly exciting.</p><p>That’s the point.</p><p>I didn’t write TARGET(AI) because AI needs help typing code. AI is getting very good at typing code.</p><p>I wrote it because a capable agent can move very fast in the wrong direction when the assumptions underneath its work are wrong.</p><p>This time, the feedback loop caught one.</p><p>I’m now trying to send the fix back to OpenCode. Maybe it gets accepted. Maybe it doesn’t. That’s up to the maintainers.</p><p>But the experiment already gave me something I wanted.</p><p>I had a good plan.</p><p>Then I found out it was wrong.</p><p>That’s what the loop was for.</p><img src="https://medium.com/_/stat?event=post.clientViewed&referrerSource=full_rss&postId=fb8a6ab2a763" width="1" height="1" alt="">]]></content:encoded>
        </item>
        <item>
            <title><![CDATA[I Wrote a Method for Goals. Then I Applied It to an AI project]]></title>
            <link>https://medium.com/@matheusreis_13304/i-wrote-a-method-for-goals-then-i-applied-it-to-an-ai-project-7aae47b224bf?source=rss-2da90e192759------2</link>
            <guid isPermaLink="false">https://medium.com/p/7aae47b224bf</guid>
            <dc:creator><![CDATA[Matheus Reis]]></dc:creator>
            <pubDate>Mon, 31 Aug 2026 12:28:28 GMT</pubDate>
            <atom:updated>2026-08-31T12:28:28.238Z</atom:updated>
            <content:encoded><![CDATA[<p>Note:</p><blockquote><em>I’m looking for 10 volunteers. To get </em><strong><em>2 free books. No strings attached</em></strong><em>.</em></blockquote><blockquote><em>I recently published TARGET, a short book about a method I developed for working toward goals when the path isn’t clear and conditions keep changing.</em></blockquote><blockquote><em>Now I want to know how it reads outside my own head.</em></blockquote><blockquote><em>I’ll send a free copy to </em><strong><em>10 people</em></strong><em> who are genuinely interested in reading it. No review, post, recommendation, or promotion expected. No strings attached.</em></blockquote><blockquote><em>I’d only ask for a few lines of </em><strong><em>private feedback to me </em></strong><em>when you finish: what stayed with you, what wasn’t clear, and whether you could see yourself applying the method to something real.</em></blockquote><blockquote><em>That’s it. Reinforcing: no need to write anything on any channel. </em><strong><em>I just want the feedback to myself.</em></strong></blockquote><blockquote><em>If you’re interested, send me a message (via DM in Medium or directly to my email: matheusgr.pmp@gmail.com) . I’ll send the book to the first </em><strong><em>10 volunteers</em></strong><em>.</em></blockquote><blockquote><em>And if one of those 10 ends up TARGETing something, even better.</em></blockquote><p>A few weeks ago, I published a small book called <em>TARGET</em>. It started with a problem that had bothered me for years.</p><p>We set goals. Then we make plans. The plan gives us comfort. It turns uncertainty into steps, dates, milestones, and checklists.</p><p>Then reality starts moving. Conditions change. Assumptions fail. New information appears. The plan grows old while we are still executing it.</p><p>I wanted a method built for that reality. That became TARGET.</p><h3>The basic idea</h3><p>A goal gives you a destination. Your current state gives you a position. Between them sits a gap.</p><p>You reason about that gap, choose an action, observe what happens, learn from the result, and adjust. Then you repeat.</p><p>Think of an aircraft crossing the Atlantic. The crew knows the destination. The navigation system knows the aircraft’s position. Weather moves, winds change, routes close, and conditions evolve during the flight. The aircraft keeps measuring and correcting.</p><p>Progress comes from thousands of small corrections.</p><p>TARGET applies the same principle to goal-directed action. The method uses feedback at three levels. One loop adjusts execution. Another examines the reasoning behind the action. A third asks whether the larger direction still makes sense.</p><p>That structure became the first book: <em>TARGET: A General Method for Goal-Directed Action, Learning, and Adaptation.</em></p><p><a href="https://www.amazon.com/dp/B0HGGT666T">Amazon.com: TARGET: A General Method for Goal-Directed Action, Learning, and Adaptation eBook : Reis, Matheus Guilherme: Kindle Store</a></p><p>Then I applied it to AI.</p><h3>TARGET(AI)</h3><figure><img alt="" src="https://cdn-images-1.medium.com/max/1023/1*mJhm3LD_T4--5XqEer-E4Q.jpeg" /></figure><p>I work with software and spend a great deal of time thinking about AI systems. Agentic AI has an interesting problem.</p><p>Agents can plan, use tools, delegate work, inspect results, and try again. Give them enough autonomy and another question appears:</p><p><strong>How should an agent decide what to do next?</strong></p><p>That question felt familiar.</p><p>An agent has a target and a current state. It acts. The environment responds. The agent observes the result and needs to decide whether to continue, adjust its action, reconsider its reasoning, or question the strategy itself.</p><p>That looked remarkably close to the structure I had already developed.</p><p>So I ran an experiment. I took TARGET and applied it to agentic AI. The exercise became much larger than I expected and eventually became a second book: <em>TARGET(AI).</em></p><p><a href="https://www.amazon.com/dp/B0HH8J89J8">Amazon.com: TARGET(AI): A Method for Goal-Directed Action, Tested Against AI and Autonomy eBook : Reis, Matheus G.: Kindle Store</a></p><h3>TARGET(X)</h3><p>I had written TARGET as a general method. Applying it to AI gave me a way to test that claim.</p><p>The core method stayed stable while the application changed. That led to a simple architecture:</p><p><strong>TARGET is the method. TARGET(X) is the method applied to a domain.</strong></p><p>AI became the first X. Other people may find very different ones: a project, a career decision, a learning problem, a business, a personal system, or any difficult goal with an uncertain path.</p><p>That is the experiment I want to run now.</p><h3>I need readers more than buyers</h3><p>Of course I want people to buy the books. Every author does. At this stage, I am more interested in what happens after someone reads them.</p><p>Does TARGET make sense outside my own head? Can someone apply it without me explaining it? Where does the method become unclear? Where does it fail? Which parts survive contact with a real problem?</p><p>Those answers matter. A framework becomes interesting when other people can use it.</p><p>So this is my invitation.</p><p>Read TARGET if the problem sounds familiar. Apply it to something real. Small is fine.</p><p>Give it a target. Establish where you are. Reason about the gap. Act. Observe what happens. Adjust.</p><p>Then tell me what happened.</p><p>Especially tell me where it failed.</p><p>I already know what TARGET looks like when I use it. Now I want to see what happens when other people do.</p><p><strong>Name the your “X”. </strong>A personal challegent, or even an already existing SAAS.</p><p><strong>Then TARGET(X). Everything is Targettable!</strong></p><img src="https://medium.com/_/stat?event=post.clientViewed&referrerSource=full_rss&postId=7aae47b224bf" width="1" height="1" alt="">]]></content:encoded>
        </item>
        <item>
            <title><![CDATA[The Problem With Goals Is That We Pretend We Know the Path]]></title>
            <link>https://medium.com/@matheusreis_13304/the-problem-with-goals-is-that-we-pretend-we-know-the-path-d17222c05946?source=rss-2da90e192759------2</link>
            <guid isPermaLink="false">https://medium.com/p/d17222c05946</guid>
            <category><![CDATA[continuous-improvement]]></category>
            <category><![CDATA[cycle]]></category>
            <category><![CDATA[target]]></category>
            <category><![CDATA[goals]]></category>
            <category><![CDATA[methods]]></category>
            <dc:creator><![CDATA[Matheus Reis]]></dc:creator>
            <pubDate>Wed, 26 Aug 2026 15:30:28 GMT</pubDate>
            <atom:updated>2026-08-28T15:31:53.113Z</atom:updated>
            <content:encoded><![CDATA[<figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*6NwEzJoVASvRV4ISW9xUog.png" /><figcaption>TARGET: what it is</figcaption></figure><p>TARGET started with a simple question:</p><p><strong>How do you keep moving toward an objective when the path is uncertain?</strong></p><p>That question took me through several established ways of thinking about decisions, execution and adaptation. And also thru decades to figure out.</p><p>The <strong>Military Decision-Making Process (MDMP)</strong> caught my attention because of its clarity around the desired end state, current situation, courses of action and execution. I have spent a fair share of my early life in the Military.</p><p><strong>OODA</strong> added something important: observation and adaptation happen continuously while events unfold.</p><p><strong>Feedback loops</strong> provided the connection between action, measurement and correction.</p><p><strong>Systems thinking</strong> helped explain why changing one part of a problem can produce effects somewhere else.</p><p><strong>Continuous improvement and experimentation</strong> added another useful principle: treat action as a source of evidence and use the results to improve the next attempt.</p><p>Each gave me something useful. I also kept running into situations where I wanted pieces from several of them at once.</p><p>MDMP was developed for military planning. OODA is powerful for rapid adaptation and competitive environments. Continuous improvement tends to assume an existing process to improve. Goal-setting frameworks are good at defining destinations, while often saying less about what happens after reality invalidates the original plan.</p><p>My own needs crossed those boundaries, unfortunately. So, I started cherry picking, adapting, but not effective nor reproducible.</p><p>I wanted one small method I could use for a project, learning something difficult, making a decision, pursuing a personal objective, or even structuring the behavior of an AI system.</p><p>That search eventually became <strong>TARGET</strong>.</p><h3>A plan is a hypothesis</h3><p>Planning works extremely well when the path is reasonably understood.</p><p>Many worthwhile objectives have much more uncertainty.</p><p>You might be learning a difficult skill, improving a project, changing professional direction, building something new, or solving a problem you only partially understand.</p><p>The destination can be clear while the route remains uncertain.</p><p>This changed the way I started looking at plans:</p><p><strong>A plan is a hypothesis about what should work.</strong></p><p>Action tests that hypothesis.</p><p>Reality produces evidence.</p><p>The evidence informs the next decision.</p><p>The resulting cycle looks roughly like this:</p><p><strong>Objective → Reason → Act → Track → Assess → Adjust → Repeat</strong></p><p>Every cycle generates information that was unavailable when the original plan was created.</p><p>That information has value.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*j5o22KDVz2BE5lUguGoXyQ.png" /></figure><h3>The framework emerged through use</h3><p>The pieces gradually started connecting.</p><p>MDMP gave me the importance of the desired <strong>end state</strong> and understanding the <strong>current situation</strong>.</p><p>OODA reinforced speed of <strong>observation and adaptation</strong>.</p><p>Feedback-loop thinking made <strong>measurement </strong>essential.</p><p><strong>Experimentation </strong>suggested treating actions as tests rather than commitments to being right.</p><p>Systems thinking kept the <strong>broader environment</strong> visible.</p><p>Eventually I found myself applying the same underlying logic to very different situations.</p><ul><li>Define where I want to go.</li><li>Understand the current state.</li><li>Reason about what might move things forward.</li><li>Act.</li><li>Track what actually happened.</li><li>Assess the difference between expectation and reality.</li><li>Use that information in the next cycle.</li></ul><p>I wanted this logic to be small enough to remember and structured enough to use repeatedly.</p><p>That became <strong>TARGET</strong>.</p><h3>Goal-directed action under uncertainty</h3><p>TARGET is a general method for goal-directed action, learning and adaptation.</p><p>Its central concern is the quality of the next decision.</p><p>Consider learning something new.</p><p>You could create a detailed six-month curriculum on the first day. At that point, however, you have very little evidence about where your actual difficulties will be.</p><p>An <strong>adaptive approach</strong> starts with an objective and an initial path.</p><p>Study. Practice. Build something. Track where progress slows down.</p><p>Perhaps the material you expected to be difficult turns out to be easy. Something you barely considered becomes the real obstacle.</p><p>That discovery changes what you do next.</p><p>The destination may stay exactly where it was.</p><p><strong>Your map gets better.</strong></p><p>That principle extends far beyond learning.</p><p>Projects generate new information. Decisions reveal consequences. Experiments invalidate assumptions. Environments change.</p><p>Progress becomes a sequence of increasingly informed decisions rather than faithful execution of an increasingly outdated plan.</p><h3>TARGET becomes an experiment</h3><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*aGoyk8rNrsP6m6Ai__ap5A.png" /></figure><p>Eventually the framework became coherent enough that I wanted to formalize it.</p><p>The result is <a href="https://www.amazon.com/dp/B0HGGT666T"><strong>TARGET: A General Method for Goal-Directed Action, Learning, and Adaptation</strong></a>.</p><p>TARGET draws from ideas with long histories: feedback systems, adaptive planning, experimentation, decision-making and learning from evidence.</p><p>My contribution is the <strong>synthesis</strong>: <em>a compact operating method designed to make those principles usable across different kinds of objectives.</em></p><p>And now comes a particularly interesting part of the process.</p><p>Until recently, TARGET had essentially one user: me. And boy, I use it everyday.</p><p>Publishing the book puts the framework into other people’s hands.</p><p>People will interpret it differently. They may find applications I never considered. Some parts may prove more useful than others. Weaknesses may become visible.</p><p>All of that becomes evidence. Data that can be ingested and analyzed under TARGET.</p><p>So the publication of TARGET is also a TARGET cycle.</p><p>I formed a hypothesis that the method could be useful beyond my own thinking.</p><p>I formalized it.</p><p>I published it.</p><p>Now I can observe what happens.</p><p><strong>Track. Assess. Learn. Adjust.</strong></p><p>Then run the next loop.</p><img src="https://medium.com/_/stat?event=post.clientViewed&referrerSource=full_rss&postId=d17222c05946" width="1" height="1" alt="">]]></content:encoded>
        </item>
        <item>
            <title><![CDATA[Quick History of Nuclear electricity generation in Brazil]]></title>
            <link>https://medium.com/@matheusreis_13304/quick-history-of-nuclear-electricity-generation-in-brazil-a3e943a353fa?source=rss-2da90e192759------2</link>
            <guid isPermaLink="false">https://medium.com/p/a3e943a353fa</guid>
            <dc:creator><![CDATA[Matheus Reis]]></dc:creator>
            <pubDate>Sat, 10 Aug 2024 16:39:58 GMT</pubDate>
            <atom:updated>2024-08-10T16:39:58.718Z</atom:updated>
            <content:encoded><![CDATA[<h3>The Brazil-Germany Collaboration</h3><p>In the mid-1970s, Brazil faced the challenge of meeting its growing energy demands while minimizing environmental impact. Recognizing the potential of nuclear power, Brazil sought international partnerships to develop its nuclear energy sector. One of the most significant collaborations was with West Germany (now unified as Germany).</p><ol><li>Agreement and Technology Transfer:</li></ol><p>In 1975, Brazil and West Germany signed an agreement that laid the foundation for cooperation in nuclear energy. The central goal was to enhance Brazil’s nuclear capabilities by leveraging German expertise and technology.</p><p>As part of this agreement, West Germany committed to supplying Brazil with eight pressurized water reactors (PWRs), each with a capacity of 1,300 megawatts (MW). These reactors were to be installed over a 15-year period.</p><p>2. The Angra Nuclear Power Plants:</p><p>The Angra Nuclear Power Plant (Central Nuclear Almirante Álvaro Alberto, or CNAAA) became the focal point of this collaboration. Located in Angra dos Reis, Rio de Janeiro, it was designed to house the German-designed PWRs.</p><p>The first reactor, Angra 1, began commercial operation in 1985. It was a milestone for Brazil’s nuclear energy program, as it marked the country’s entry into nuclear electricity generation.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/0*l2IKAE33vO74LnTg" /></figure><p>When Angra 1 kicked off in 1985</p><p>3. Angra 2 and Angra 3:</p><p>Angra 2, another PWR, followed suit and started operating in 2001. Its construction faced delays due to technical challenges and financial constraints.</p><p>Angra 3, initially planned as part of the original agreement, experienced further delays and was suspended at various points. As of now, its completion remains uncertain.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/0*kgY5pwt1Vp7ZCeKn" /></figure><p>The kick off of Angra 2 more than doubled the share of nuclear energy consumption in Brazil. Angra 3 still halted due to bureocracy/other discussions since 2015.</p><blockquote><em>Key takeaway 1: the Brazil-Germany partnership laid the groundwork for nuclear electricity generation in Brazil, with Angra Nuclear Power Plant serving as a testament to this collaboration. The journey has been marked by challenges, but it underscores the importance of international cooperation in advancing sustainable energy solutions.</em></blockquote><blockquote><em>Key takeaway 2: Brazil´s electricity generated by nuclear plants remains at ~2% of the whole country´s powerplants, which is a very low contribution despite the ever need for more clean energy.</em></blockquote><h3>Note about Germany:</h3><p>Germany’s share of nuclear energy for electricity dropped significantly due to a deliberate decision to phase out nuclear power.</p><p>1. Anti-Nuclear Movement</p><p>2. Renewable Energy Advocacy</p><p>3. Green Party Influence</p><p>4. Energiewende (Energy Transition)</p><p>5. Recent Shutdown:</p><p>- On April 15, 2023, Germany shut down its last nuclear power plant, prioritizing safety and climate action.</p><blockquote><em>Germany’s shift away from nuclear power reflects a complex interplay of historical context, environmental concerns, and energy policy decisions. The focus now is on renewables and a sustainable energy future.</em></blockquote><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/0*R-pymkTlvTnmlojU" /></figure><p>The share of nuclear energy production in Germany dropped gradually from 32% in 1997 to less than 3% in 2023.</p><p>Sources:</p><p>[1] Kaggle: <a href="https://www.kaggle.com/datasets/alistairking/nuclear-energy-datasets">Nuclear Energy Datasets (</a><a href="http://kaggle.com/">kaggle.com</a><a href="https://www.kaggle.com/datasets/alistairking/nuclear-energy-datasets">)</a></p><p>[2] Microsoft Copilot — Text generation</p><p>[3] PowerBI: Data cleaning and data visualizations</p><img src="https://medium.com/_/stat?event=post.clientViewed&referrerSource=full_rss&postId=a3e943a353fa" width="1" height="1" alt="">]]></content:encoded>
        </item>
        <item>
            <title><![CDATA[What impacts student academic success in tests]]></title>
            <link>https://medium.com/@matheusreis_13304/what-impacts-student-academic-success-in-tests-624ee41bc77c?source=rss-2da90e192759------2</link>
            <guid isPermaLink="false">https://medium.com/p/624ee41bc77c</guid>
            <dc:creator><![CDATA[Matheus Reis]]></dc:creator>
            <pubDate>Sat, 10 Aug 2024 16:22:25 GMT</pubDate>
            <atom:updated>2024-08-10T16:22:25.247Z</atom:updated>
            <content:encoded><![CDATA[<h3>Objective:</h3><p>This is another project focusing on variables that impact student performance in tests. Specifically, we want to understand the influence of the parents background, test preparation and others on students performance.</p><h3>Background:</h3><p>Data analysis plays a vital role in enhancing educational outcomes. In this project, you’ll utilize a student performance dataset containing grades, attendance records, and demographic details. By using Excel, we will examine trends in student performance, pinpoint factors affecting academic achievement, and suggest improvement strategies. This project showcases the need to use data for informed decision-making in the education field.</p><p>Source of data: <a href="https://www.kaggle.com/datasets/spscientist/students-performance-in-exams">Kaggle</a></p><h3>Initial inspection:</h3><p>It´s a simple dataset already in .xlsx format, requiring no special data transformation. However, as we know, first thing is to load the data into powerquery to visually inspect the health of the data:</p><p>1 — Open the file in Excel and, on Data Tab, click on From Table/Range:</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/0*8KrcomxIynC0mQn0" /></figure><p>How to open powerquery from an already existing table</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/0*Eyg9imsX8eeJpzhs" /></figure><p>Columns have values, no errors and data types are correct. Close and load.</p><h3>Exploratory Data analysis:</h3><p>This dataset, per se, is pretty straight forward and might seem — at first look — dull to work with. No challenges at all, if your goal is to use complicated formulas and show off your excel skills.</p><p>That should never be the goal of any serious data analysis.</p><blockquote><em>Your job is to find the essence, the the story the data is telling you. You are a detective, looking for clues and likely hypothesys to explain a situation. As a result of this, you will be able to take actionable steps in the future.</em></blockquote><p>Let´s begin.</p><h3>Scores distribution shapes:</h3><p>For this analysis, let´s first take a look at each of the 3 scores shapes: do we have normal distributions, binomial, skewed, etc?</p><p>I´m using histogram charts this time, but we could use the data analysis pack and see the descriptive statistics. Your choice.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/0*HH4NHRERkYAk09zB" /></figure><p>Bins are configurable. The more bins, the more granularity in your shape.</p><p>Also, in this dataset, there are more females than males:</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/750/0*q6b8vTpXWV6ORZ0f" /></figure><p>Males x females in dataset. Do not truncate Y axis!</p><blockquote><em>Important: To shield us from a very common trick to lying with charts, see chart below: Take a look at this biased representation. Be aware of TRUNCATED Y AXIS!!!</em></blockquote><figure><img alt="" src="https://cdn-images-1.medium.com/max/750/0*RnIqWcbuqzJhkIIG" /></figure><p>Lying with charts: be aware of Y axis truncated. The intention is to visually deceit us to consider a significant difference where it was very low in fact (7% less males only)</p><h3>Gender x Scores:</h3><p>This is an interesting topic, and for that I will use a boxplot chart. Before looking at the chart, I recommend you recap what a bloxplot is in <a href="https://datatab.net/tutorial/box-plot">this very good explanation</a> by DataTab.</p><p>Why boxplot: it´s just because it´s a powerful and condensed way of seeing data dispersion, it´s mean, median and quartiles (if you divide a dataset in 4 parts, each part is 1 quartile = 25%)</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/817/0*4X6On2EXqgtdch5d" /></figure><p>From <a href="https://datatab.net/tutorial/box-plot">Box Plot — Simply explained — DATAtab</a></p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/0*6R73QfUgoh76TLdz" /></figure><p>In this dataset: males performed</p><h3>Preparation course attended x Scores:</h3><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/0*v9rbMveXdRraeADa" /></figure><h3>Parents´ level of education x Scores:</h3><p>That´s in interesting topic, and using this dataset here´s what we find out:</p><blockquote><em>the highest the education level of the parents, the better the students perform in all 3 scores.</em></blockquote><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/0*VuCX19hZU8n38a6Z" /></figure><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/0*_7WJkRApANvxW-SI" /></figure><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/0*g9Cs-YFsLHCvoV1g" /></figure><p>Note:</p><ul><li>Bachelor’s Degree: An undergraduate degree awarded after completing a four-year college or university program.</li><li>College Degree: A general term for any degree earned from a college, including associate, bachelor’s, master’s, or doctoral degrees.</li><li>Master’s Degree: A graduate degree awarded after completing a specialized program beyond a bachelor’s degree.</li><li>Associate Degree: An undergraduate degree typically earned after two years of study at a community college or technical school.</li><li>High School Degree: A diploma awarded upon completing the required coursework and exams in high school.</li><li>Some High School: Indicates that a person attended high school but did not complete the requirements for a diploma.</li></ul><h3>Executive summary:</h3><ol><li>Students who attend preparation courses do better in all 3 tests;</li><li>The higher the education level of the parents, the better the perfomance in the test</li><li>Gender scores: this needs some considerations. Althought this dataset shows men performing slightly better in math, and women performing better in reading and writting, specially with math there´s no significant difference when subject controlled studies are done. References <a href="https://www.scientificamerican.com/article/are-boys-better-than-girls-at-math/">here</a> and <a href="https://www.ncbi.nlm.nih.gov/pmc/articles/PMC4270278/">here</a>. We also have no idea how this dataset was taken and under what scientific methodologies.</li></ol><p>That´s it, everyone.</p><pre>This  project highlighted some handy charts in order to extract the maximum info from a dataset without the need of using statistical calculations, and finished of concluding the need of using data for informed decision-making in the education field.</pre><img src="https://medium.com/_/stat?event=post.clientViewed&referrerSource=full_rss&postId=624ee41bc77c" width="1" height="1" alt="">]]></content:encoded>
        </item>
        <item>
            <title><![CDATA[A step-by-step guide to finding patterns and trends using Excel (and a bit of statistics)]]></title>
            <link>https://medium.com/@matheusreis_13304/a-step-by-step-guide-to-finding-patterns-and-trends-using-excel-and-a-bit-of-statistics-274e938a5833?source=rss-2da90e192759------2</link>
            <guid isPermaLink="false">https://medium.com/p/274e938a5833</guid>
            <category><![CDATA[excel]]></category>
            <category><![CDATA[data-analysis]]></category>
            <category><![CDATA[data-visualization]]></category>
            <category><![CDATA[power-query]]></category>
            <dc:creator><![CDATA[Matheus Reis]]></dc:creator>
            <pubDate>Tue, 30 Jul 2024 23:25:14 GMT</pubDate>
            <atom:updated>2024-07-30T23:25:14.475Z</atom:updated>
            <content:encoded><![CDATA[<p>Humans have been designed to spot patterns everywhere, even when don´t exist. Long before walking, the <a href="https://en.wikipedia.org/wiki/Homininae">Homininaes</a> realized that a pair of dark dots nearby might mean a predator or a threat. Some took their chances and ignored the signs, others took that seriously and thanks to them we´re communicating via this post.</p><p>The downside of our brain´s radar for patterns is that we can even spot patterns where they don´t exist, which misleads us to take actions not needed. But it´s undeniable that we survived thanks to several factors, one of them being our brains´ ability to see patterns and trends.</p><p>With this brief introduction, let´s get back to one real life scenario: you´ve just been contracted to assign a financial risk level to a given person from any given country. There are several (fictional) data associated with each individual, and based on this the person is given a risk level: Slow, Medium or High.</p><p>You know squat about credit scores, financial risks and all the algorithms used to assign risk level. What do you do?</p><p>In the first week of your job, apart from all the courses and knowledge sharing sessions with your colleagues, you have an idea: what if you pulled from the company´s records the latest financial risk assessment available?</p><p>You then get this (fictional) risk assessment file (which is from <a href="http://kaggle.com/">Kaggle.com</a>) and you try to discover more…</p><h3>1 — Open the file using Power Query:</h3><p>You open it in Excel using power query, just to see if you can transform from the get go some data or spot any inconsistencies:</p><p>Initial inspection shows your columns values are ok in the majority of the cases. You fill in some missing values with “null” and you inspect other info: table contains columns with age, gender, education level, marital satus, income, credit score, loan amount, etc etc.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/0*LvH7_dleQWWj-45o" /></figure><p>Data intake via Power Query. Note: Anyone who wants to use Excel to its full capability needs to learn PowerQuery and PowerPivot. Take note of this and find WHY so.</p><h3>2 — Initial questions:</h3><p>You start to think: is there any correlation between age and financial risk? Is there any relation between my credit score and financial risk?</p><p>Those are valid questions, but a table at this point won´t answer any of them. You stop your work to have lunch but, during your lunch, your mind cannot let it go: you decide that, when you´re back from lunch, you go crazy on plotting.</p><p>You´re not being paid to do graphic design. You´re being paid to find meaningful info for informed decisions. That´s the key skill that is highly on demand in today´s job market or in your company. Think that you put your your family house as a collateral in case you setup a consulting company. You have skin in the game.</p><h3>3 — Plotting data:</h3><ul><li>Does having a good payment history affects your financial risk?</li></ul><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/0*PbDaLyLeMOwNkpDa" /></figure><p>No apparent correlation or trend can be seen with this visualization</p><ul><li>How about employment status? There might be some positive impact having a job so your financial risk being low, you reckon.</li></ul><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/0*LoN69ES4HVi9ygos" /></figure><p>The data disagrees with your hunch. Now what? You grow frustrated and blame it on Excel (we all do, sometimes)</p><p>Again slight variations, but the data shows that there´s no apparent correlation between your employment status x your risk level. You start getting puzzled, and decides to have a coffee.</p><p>You remember your goal is to find some correlation, or some pattern. But so far there´s none! What else you could do?</p><ul><li>Correlation parameters:</li></ul><p>You get tired of plotting after the coffee but bring the big guns: let´s run some stats on those series. It´s impossible to not find anything meaningful until now, you think. You proudly open the Data Analysis pack:</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/0*LfpC8-v2pQXe_1Hq" /></figure><p>Instructions on how to turn on this analysis pack can be found @office.com</p><p>Your idea is to run a couple of comparisons to find correlations numerically. Also, you want to find out if the Risk Rating column follows a “normal” distribution. You end up with this:</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/246/0*Cu1a5IuwCLlW_Sam" /></figure><p>What does this tell you? If you don´t know, go after the info.</p><p>From Stats 101, you already know what´s the meaning of mostly everything here, and your gut feeling is that is a gaussian distribution. THEN you spot the Kurtosis: If it is 0 or close, it´s a Gaussian distro. In this case, you can confidently say that it´s not the case.</p><p>By plotting the Risk Rating, you get a visual confirmation:</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/0*WT7PzMvvSB6cTZZL" /></figure><p>This chart is a Histogram. In this sense, we´re seeing ranges of values in X axis and Frequencies on the Y axis</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/0*kQhkhnWlLvXsdeGT" /></figure><p>Looks like a banner saying: It´s not a “normal” distribution</p><p>So, after having your 2nd coffee, you go back to your original idea of finding correlations (mind that Data Analysis pack only do this on numeric values, so you convert your risk ratings to 1 (low), 2 (medium) and 3 (high).</p><p>You try to see if there´s any correlation between Debt-to-Income Ration and Credit Score, and found out there´s none.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/452/0*xbfE6swS1ax56Rwq" /></figure><p>Correlation way below 1. No correlation.</p><p>You do that for a couple of columns, and find the same result. No apparent correlation.</p><h3>4 — The light at the end of the tunnel:</h3><p>To wrap up your day´s work, you try one combination that never crossed your mind: during your inspection of data in Power Query, there´s one column called “Country”. Since you are running out of ideas, you give this a shot: you compare the country with the Individual risk rating.</p><p>It makes no sense, but what you´ll lose by doing it? Just a few minutes.</p><blockquote><em>Wait up! What´s that?</em></blockquote><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/0*R_WZILTtC864Osps" /></figure><p>You finally found something workable: There are countries with individual´s Risk level significantly different. You struck gold here. You start giving Excel some credit.</p><p>By plotting in a stacked column chart, you managed to find some info: Phillipines have a higher Risk Rate compared with Aruba, for instance. You struck gold here, and the next question: Why is that?</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/0*IRkp8z9_3wBpSJ1n" /></figure><p>Data shows the Risk Rating algorithm assigns different Risk levels depending on the country. Is this a rule or a consequence?</p><p>To finish off your day, you plot the events per country, not in a stacked way:</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/0*mASbu3fHbIQ50cLw" /></figure><p>Low (green), Medium (orange), High (red)</p><p>In this fictional dataset, some of the key takeaways after some visual inspection taken by the chart above:</p><ul><li>Falkland Islands (Malvinas): Shows a significant high-risk rating.</li><li>Bulgaria: Medium risk seems to be the most frequent rating.</li><li>Congo: High risk is notably less than medium risk.</li><li>United States Minor Outlying: Predominantly low to medium risk.</li><li>Georgia: A balanced distribution across all three risk levels.</li><li>Panama: Medium risk is slightly more prevalent than low and high risks.</li><li>San Marino: Shows a considerable amount of high risk.</li><li>Niger: High risk is quite prominent.</li><li>Guinea: Medium risk is slightly more prevalent than low and high risks.</li></ul><blockquote><em>You feel proud to have understood a bit of how your company is assigning risk rates, and you definitely found out some country variations. More importantly: you have questions now!</em></blockquote><p>Next steps: go ask a bit more about the algorithm used so you can clarify WHY this is happening.</p><p>On your next day, you will continue cross checking these variations to discover more: what´s the influence of previous defaults in the risk rating?</p><p>That´s the end of the Exploratory Data Analysis on your fictional set, at least for now!</p><p>Good job.</p><img src="https://medium.com/_/stat?event=post.clientViewed&referrerSource=full_rss&postId=274e938a5833" width="1" height="1" alt="">]]></content:encoded>
        </item>
        <item>
            <title><![CDATA[Very offensive.]]></title>
            <link>https://medium.com/@matheusreis_13304/very-offensive-4fc465147a0e?source=rss-2da90e192759------2</link>
            <guid isPermaLink="false">https://medium.com/p/4fc465147a0e</guid>
            <dc:creator><![CDATA[Matheus Reis]]></dc:creator>
            <pubDate>Mon, 27 Jul 2020 18:09:58 GMT</pubDate>
            <atom:updated>2020-07-27T18:09:58.734Z</atom:updated>
            <content:encoded><![CDATA[<p>Very offensive. Calling people white/black and IMPOSING one behaviour on others based on their skin tone. That´s totally messed up, post reported as harassement.</p><img src="https://medium.com/_/stat?event=post.clientViewed&referrerSource=full_rss&postId=4fc465147a0e" width="1" height="1" alt="">]]></content:encoded>
        </item>
    </channel>
</rss>