Causion is a free web app for building causal diagrams (DAGs) and then watching them behave—by simulating data and switching from observation to intervention with do(). Try it now: https://causion.pedermisager.org/ (it runs in your browser)

In this post, we will use a practical example to explore the power of statistical control, explaining why adjusting for the right variables clarifies relationships and why adjusting for the wrong ones can introduce new, serious biases.

In this blog post, I will briefly explain why experimental designs are considered so valuable by scientists and why we often insist that randomized experiments are necessary for answering causal questions. I will use a practical example to illustrate this point. If you’re new to the concept of causal inference, it may be helpful to read my earlier blog posts on correlation vs. causation and eight basic rules for causal inference.

In this blog post I will describe eight basic rules that govern the relationship between causal mechanisms in the real world and associations/correlations we can observe in data. To make each rule as easy as possible to understand, I will describe each rule both in words and in causal graph and logic terms, and I will offer some very simple simulation R code for each rule to demonstrate how it works in practice.

In this blog post I propose a four one-sided test of significance (FOST) that can be used to test whether an effect is smaller than OR larger than OR equivalent to the smallest effect size of interest. I lay out the general setup and interpretation of FOST, and ask readers to help me evaluate its validity.