The thing about my way of teaching stats is that my cool, novel examples get old really fast. Like, pop culture is on a thirty-day cycle. Anyway, YEARS ago, 2014 to be exact, I blogged about illustrating bimodal data distributions using baby name data from the Social Security Administration. Specifically, when an older name comes back into vogue, it has a bimodal distribution when you plot time on the X axis. I stand by this example. However, all of the data illustrations come from 2014. When our traditionally aged college students were 6 or 7. Which makes this example feel a little stale. However, here is some data, courtesy of Redditor fevaah , showing old names that are back in vogue through 2025 ! And Eleanor, Violet, Nora, and Hazel are here, illustrating bimodal distributions for a new generation. If you used my old baby-name data, this is a quick, easy refresh for your lectures (or a fun new addition if you didn't). If this idea was help...
This blog post contains one small analogy for explaining family-wise error to your students. I was making French toast for dinner the other night. While I was measuring out cinnamon, I realized using one tablespoon instead of three teaspoons to avoid measuring errors is sort of like using a one-way ANOVA with three levels instead of doing three t tests to avoid Type I error. Stick with me here. If I were to use three teaspoons to measure out an ingredient, there is a chance I could make a mistake three times. Three opportunities for air pockets. Three opportunities to not perfectly level out my ingredient. Meanwhile, if I just use one tablespoon, I will only risk the error associated with using a measuring spoon once. Similarly, every time we use NHST, we accept 5% Type I error (well, if you are a psychologist and using the 5% gold standard, but I digress). Using three tests ( t tests) when we could use one (ANOVA) will increase the risk of a false positi...