A lot has been written on the reasons why researchers should learn how to program. I’d like to add one: If you are like me, nobody has taught you about the more “recent” developments in statistics, such as Bayes factors or linear mixed models. Now we are somehow expected to figure them out ourselves. This is a difficult task, but being able to program helps. I really only “got” Bayes factors after I implemented some myself in R, and there is still a lot I do not know. If you know about Monte Carlo simulations, you can try to figure out the behavior of any method under varying conditions, increasing your understanding of—and trust in—a method, even if you lack a strong mathematical background. As an example, see the simulation tutorial on linear mixed models by Lisa DeBruine and Dale Barr. I am not arguing that programming is the optimal method for learning for everyone. But I do think it often helps.
Last updated: 2019-11-29