Some notes on categorical data analysis, using simulation in Stata and R

Simple modelling of categorical data is not as simple as it seems. Standard formulas taught to generations of undergraduates are shown to be sub-optimal, simple models are widely misunderstood, and high levels of controversy surround the suitability and interpretation of relatively standard models such as logistic regression. In this research note I discuss a number of these issues, using simple simulations in Stata and R to illuminate them.