One of the most often mistaken and forgotten problem in basic statistics, when they assume that the fitted regression model is symmetric, aka. does not matter if you fit x~y or y~x. Well, based purely on the method and definition of linear model fitting, this is a very wrong assumption and I think one of the main cause for this is that linear models are always interpreted together with correlation values where this causational direction is not present. Let's see why and learn how to use and interpret correctley, morover what if you need a symmetric model.
The second episode of the Jellybean Mistakes series. Let's talk about barplots and error bars, where and how to use them appropriately and what other approaches there are for better representation of your data.
Introduction to a new series about common mistakes in statistics, datavis and everything else in biology.
Recently I got the opportunity to attend to SOKENDAI's Freshmen course. Here I summarise the questions and answers I got on my presentation.