IOED Stats

I repurpose the words of Adam Waytz who, through his literature in This Idea Is Brilliant, planted the seed for this space.

If you asked 100 data analysts/scientists/[plug your title] whether they understand what confidence intervals are, most would say yes. But ask them to produce a detailed, step-wise explanation of exactly how it works, and you’d likely hear stammering or empty words. This powerful but inaccurate feeling of knowing is what Leonid Rozenblit and Frank Keil termed the illusion of explanatory depth.

Why does IOED exist in the field of statistics (alongside mathematics and data science)?

As the speed of delivering new data professionals grows, institutions are spending lesser time on strengthening the understanding of basic statistics (and mathematics and data science) terminologies they preach. There simply isn’t enough patience to toil through the foundations. After all, there are exciting ML & AI models to get to.

Another contributor is the high barrier of scholasticism - the high birth rate of technical terminology that few experts have tried to humanize (explain to the lay person), and fewer yet have given it a worthy stage (distribution channel). Prescribed definitions and ‘rules of thumb’ have replaced understanding, such that this fragmented knowledge is no longer transformed into wisdom in its bearers. As if these defintions & rules of thumbs are set in stone - when it is, in fact, the opposite. Rules in the field of Statistics (or Mathematics or Data Science) are situational (subject to a set of conditions). Understanding a method or rule involves understanding these conditions & adapting the prescribed solution accordingly. Without a strong foundational knowledge of these basics, the new age data professionals lose their ability to reason or intellectually defend a choice of analysis and its sub-methods.

How often have you questioned whether 95% confidence level is best for your model? Why 95%? If your answer is ‘because a book/article/person said so’, this wiki is for you. If your answer is ‘yes, of course, I question these paramaters’, then join me in expanding this wiki for those who’ve forgotten to question the defaults but would absorb relatable explanations if only a directory was available.

Note: This wiki is meant for everyone, and not only data practicioners, interested in building an intuitive understanding of statistical terms, methods & rules.