In Uncertainty Wednesday I have talked a lot about the danger of doing statistics without a model. Another big danger is ignoring dynamic processes and conducting static analyses instead. That is you may have an explanation, such as “a person’s height is influenced by their genes” and you will wind up drawing very different conclusions if you take a static perspective.

Here is a great chart from Our World in Data on the average height in three countries over time (a dynamic version available here)

We see a dramatic change in average height over time. The chart also shows that differences between groups sometimes persist for long periods (e.g. France compared to Germany) and sometimes fluctuate (e.g. Russia compared to France).

So there is a huge difference in how much genetic factors influence height given roughly the same environment (a moment in time) versus taking significant environmental changes into account (longitudinal). Now height happens to be something we can measure easily. Imagine how much more problematic this becomes when trying to measure something like intelligence. Our information environment has changed massively over the last couple of decades. What the brain is exposed to today and how it can be exercised (think adaptive learning) has changed dramatically. Any point in time analysis or even one that looks at a period in which the information environment has been relatively static is likely to dramatically underestimate environmental impact.