Logistic Regression and Generalized Linear Models

Chapter 6 provides an overview of logistic regression and generalized linear models (GLMs) as extensions of ordinary linear regression (OLS) for non-continuous outcomes. It explains how logistic regression handles binary outcomes by modeling the probability of an event through the logistic function, thereby constraining predictions between 0 and 1. The chapter highlights the interpretation of logistic regression coefficients via odds ratios, illustrating how predictors influence the odds of an outcome (e.g., factors doubling the odds of employee retention) under control of other variables. Model fit assessment in logistic regression is addressed using pseudo-R² measures (such as McFadden’s R² and others) and goodness-of-fit tests, noting their different interpretation compared to OLS R². The chapter also introduces the broader class of GLMs, which extend the linear modeling framework to other distributions (e.g., count or ordinal outcomes) by choosing appropriate link functions and error structures. Examples demonstrate how GLMs empower researchers to model diverse types of data commonly encountered in the digital landscape, from predicting user churn (binary outcome) to modeling counts of online engagements, while emphasizing the importance of checking model assumptions and using correct link functions for valid inference.