I think that interactive visualizations are better than lengthy, wordy text books when it comes to illustrating statistical principles. One little GIF or interactive website can do a far better job than text or words. For example: Everything that Kristoffer Magnusson has given us (effect sizes, correlations, etc.). Here is a new tool for explaining critical regions in Intro Stats. Matthew  McBee created an interactive in shinyapps that shows how critical regions change a) depending on test, b) sample size, change of the shape of the distribution. With the ol' t-test, you can show how the critical values move around with degrees of freedom What your t-test critical values looks like at df = 3... ...versus how the those critical values look at df = 80 Also, you can do the same thing but with F curves. Andplusalso: Matt has also created shiny apps to adjust p-values for multiple comparisons , AND another one for calculating p-values based on a test statistics ...

There are a lot of lessons in this one study  (Craig, Nelson, & Dixson, 2019): Within subject design, factorial ANOVA and interactions,and data is available via OSF. Let's begin: TL: DR: The original study looked and the presence or absence of beards and whether or not this affected participants' ability to decode the emotional expression on a man's face. Or, more eloquently: TL: DR: Their stimuli were pictures of the same dudes with and without beards. And those weren't just any dudes, they had been trained in the Ekman facial coding system as to make distinct expressions. Or... One participant, rating the same man in Bearded vs. Non-bearded condition, provides a clear example of within subject research design. This article also provides examples of interactions and two-way ANOVA. Here look at aggression ratings for expressing (happy v. angry) and face hairiness (clean-shaven v. beard). Look at that bearded face interaction! Bearded guy...

Eric Roston and Blacki Migliozzi put together some great data visualizations illustrating different factors that may or may not contribute to global climate change ( "What's Really Warming the World?" ). I couldn't capture it in this blog post, but the data is animated and interactive as to highlight change over time. Very slick. This got me thinking about multiple regression, which studies different variables (X 1 , X 2... ) that may or may not contribute to some outcome (Y), and how we can use this website as a conceptual example of multiple regression. Here, the graph features multiple "predictor"/X 1 , X 2 , X 3 , X 4 variables (greenhouse gasses, ozone, land use, aerosols) as well as the predicted/Y variable (global temperature). we can see the aerosols are likely a very poor predictor while greenhouse gasses are likely a good predictor. This page can also be used to explain plain old linear regression. This example compares one predictor/X...