Lecture
This lecture delves into special examples of Generalized Linear Models (GLM), focusing on logistic regression for binary data and loglinear regression for count data. The instructor explains the concept of scale parameter and the intuition behind GLM link functions. The lecture covers topics like sparseness in design matrices, separation issues in logistic regression, and the use of jittered residuals for binary responses. Additionally, it explores the challenges of perfect separation and the implications for MLE existence. The lecture concludes with a discussion on count data models, including Poisson and multinomial distributions, contingency tables, and nonparametric relationships with covariates.