11 Generalized Linear Models (GLM)

Lecture slides: GLM

Today’s exercise: Web Excercise 11

Linear models are extremely versatile, but because they are basically lines, they can predict values smaller than 0 and larger than one. For some outcomes that makes no sense. For instance, what does it mean to have a probablility 0f -1 to be dead; to have several lifes like a cat? Generalized linear models, take the predictions deriving from a linear model and project them onto the meaningful space of an outcome. For binary data that are values between 0 and 1, for count data those are values between 0 and \(+\infty\). In this session you will get a brief introduction to these models and how to estimate them in R.

To get an impression, how important GLM’s are, you might want to check out this video, which explains how epidemics follow a logisitc curve - just like the type of regression you learn in this class: