Lab 4-Binary Regression

XUEFENG conducted a lab session on binary regression, covering logit and probit models for Chapter 5. The session included instructions on loading necessary packages, particularly MFX for computing average partial effects, and using the MROZ dataset containing labor force participation data for 753 married women. XUEFENG demonstrated the problems with using ordinary least squares (OLS) for binary outcomes, showing how linear probability model predictions fell outside the 0-1 probability range. The lab covered how to estimate logit and probit models using the GLM function, interpret coefficients through odds ratios, compute average partial effects (APE), and evaluate model fit using percentage correctly predicted and pseudo R-square metrics. XUEFENG provided guidance on interpreting odds ratios and explained that pseudo R-square values are typically lower than OLS R-square values but should be around 0.2-0.4 for logit or probit models. The session concluded with instructions for an after-lab practice assignment and a separate lab assignment on binary choice, logit and probit models, which requires students to work independently with provided R code and data.