This document provides an overview of generalized linear models (GLMs) and maximum likelihood estimation (MLE).
It discusses the exponential family distribution framework for GLMs, which allows the use of the same tools of inference across different distributions. It presents examples of link functions and canonical forms for the Gaussian and Poisson distributions.
The document also covers likelihood theory and how to calculate parameter estimates and their uncertainty through taking derivatives of the log-likelihood function. It introduces MLE as a method for finding the parameter values that maximize the likelihood of observing the data. Computational estimation of GLMs is performed through an iterative least squares method using weights from the distributions' Fisher information.