ABSTRACT
Marginal standardization can convert logistic regression results into risk ratios or risk differences. These methods can extract a rate difference from a rate ratio model, along with confidence intervals, P-values, and other statistics. The recent addition of many marginal commands to Stata makes all of this easy. A few pitfalls are noted. Marginal commands can deal with interactions and with continuous variables. Thanks to marginal methods, we can use the model that has better fit to the data, multiplicative or additive, and extract rate difference or ratio estimates.
