ABSTRACT
In Chapter 3 we presented an overview of how to construct a maximum likelihood function for linear regression. A fairly complete model was presented, with code to demonstrate how to maximize the log-likelihood, determine residuals, display a summary function, and allow for post-estimation statistics. The IRLS algorithm, which provides maximum likelihood estimates for a limited set of models, was discussed in Chapter 4. We constructed several different types of functions for estimating standard GLM models, and then developed a modular umbrella irls function for estimation of binomial, Poisson, negative binomial, gamma, and inverse Gaussian models. All models discussed in Chapter 4 have only a single parameter to be estimated.
