by
Rebecca Raciborski, PhD
Seminar date: 6/21/2023
Description: The challenges of interpreting coefficients from nonlinear models (e.g., logistic models) are increasingly becoming known to health services researchers. A popular solution is to report marginal effects instead of transformations of the coefficients (e.g., odds ratios). However, other alternatives exist and may be preferable depending on the research question. This talk will discuss important nuances of marginal effects and introduce predictive margins as a tool for interpreting regression output.
Intended Audience: This talk is aimed at health services researchers who want to clearly communicate results from regression models; no prior knowledge of marginal effects is required but familiarity with nonlinear models will be helpful.
DOWNLOAD: PDF handout | Audio only (mp3) | transcript
