Skip to main content

Posts

Showing posts from September, 2025

Percentiles, bee swarm plots, Bureau of Labor Statistics data...so many lessons in one interactive chart.

 There are so many ways to use this tool: Nathan Yau's Flowing Data is one of those websites I check every few days for statistical inspiration. He shares  the work of others and his own, including this  interactive bee swarm plot that illustrates salaries  for various  jobs. The bee plot, with the cursor of Psychology Teachers. https://flowingdata.com/2025/09/09/salary-and-occupation-2024/ There are many ways to use this in stats class: 1. At some point, you should talk about career exploration with your students.   2. Statistics students should be learning about modern data visualizations like this jitter plot, aka bee swarm plot.  3. If you cursor over any dot, you can see the 25th and 75th percentile scores and n size for that occupation's salary. 4. The size of each circle corresponds to the n size. Which I love because jitter plots do a great job of illustrating variability in a data set. However, each data point here represents an average...

Teaspoons, Tablespoons, and a new analogy for family-wise error.

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...