The Gentlest of Introductions to Bayesian Data Analysis
How to think in a natural way based on data
If you have ever taken a statistics class at university or an online course, you might recall terms such as hypothesis testing, confidence intervals, or p-values. If so, what you were taught was the so-called “classical” or “frequentist” statistics, which is the dominant approach in teaching STATS101. There is, however, another school called Bayesian statistics, which builds up a different way of thinking about data and models.
The Bayesian approach is often neglected at universities and online courses alike as harder to explain, understand, and apply. I believe such branding is unjust. Actually, I think the Bayesian way of thinking is more natural and offers significant advantages over the classical approach. Let me do my best to offer you the gentlest of introductions to Bayesian statistics in this article. Let’s dive in!
Updating beliefs with data
Imagine we are at a pool table. You are standing with your back to the table so that you can’t see it. I place a white ball somewhere on the table and ask you where it is: in the right part of the table, in the left part, or in the middle? Naturally, you have no idea whatsoever and all you can do is guess. Then, I start placing color balls randomly on the table and each time I do, I tell you if the new ball is to the left or to the right of the white one. After 5 rounds, the table (which you still cannot see) may look somewhat like this: