How a simple simulation replaces common statistical tests - dr. Ofer Engel

Dr. Ofer Engel will give us a workshop on the simulation-based inference process.

Statistical tests are based on the idea of comparing features of our data (its mean, median etc...) to the features of a sampling distribution. This comparison tells us whether our data is consistent with our hypotheses.

But the sampling distribution is a theoretical and abstract concept, one that we never get to explore directly, and is therefore difficult to grasp intuitively. Moreover, there are many dozens of statistical tests in existance already, and remembering when to use which can be tricky for the uninitiated.

In this lecture I will argue that you need to know only one test. We develop the simulation-based inference process, thereby demonstrating a direct connection between data production and the logic of inference. Simulations allow us to do more with less: more ideas, less memoization, more intuition, less abstraction. Instead of analytical methods, we will use the computer to show that statistics just makes sense. Before attending the lecture, students are encouraged to watch the following video: Statistics without the agonizing pain.