MATH5806 — Applied Regression Analysis at UNSW

I took MATH5806 in 2026 T2 as part of my Masters of Statistics. The name undersells it: “applied regression analysis” here means generalised linear models, done properly — exponential families, link functions, iteratively reweighted least squares, deviance — with ordinary linear regression as the special case rather than the destination, and splines, Poisson and binomial regression along the way. Everything runs in R. This page is the admin, the textbook, and my notes, in one place.

what the course actually covers#

The arc is the classical GLM story:

1. the exponential family and its mean–variance structure
2. estimation and inference — score, Fisher information, Newton–Raphson and Fisher scoring, Wald and likelihood-ratio tests
3. normal linear models, then the generalisation: link functions and IRLS
4. binomial and logistic regression, odds ratios, separation
5. Poisson regression and log-linear models for counts
6. deviance, residuals, and model checking
7. smoothing: splines and flexible regression

I wrote the theory up as standalone wiki explainers while studying; they are deliberately interpretation-first rather than formula-dump:

• regression — the section index
• generalised linear models — exponential family, links, IRLS, deviance
• estimation & inference — MLE, score, Fisher information, Wald vs LRT
• logistic regression — developed honestly as a GLM, with IRLS from scratch in R

textbook, and worked solutions to all of it#

The course textbook is An Introduction to Generalized Linear Models (Dobson & Barnett, 4th edition). It is short, readable, and the lectures track it closely — buy or borrow it, and actually do its exercises.

While preparing for the course I worked solutions to all 78 exercises in the book, in R — exponential family theory through to contingency tables, survival analysis, longitudinal data, and Bayesian analysis with MCMC. If you are stuck on a Dobson exercise, or want to check your working, that page is the densest MATH5806 study resource on this site.

assessment#

The 2026 T2 offering ran as two lectures a week (a 2-hour and a 1-hour) plus one tutorial, assessed as:

Check the current course outline before relying on this — weights and formats drift between offerings.

Two practical notes. First, 45% of your mark is settled before the final, and the week-4 quiz arrives fast — have the exponential-family and link-function material cold early. Second, the assignment is a genuine R data-analysis exercise, not a coding formality: fitting, diagnostics, and interpretation all get marked, so the habit of reading summary() output critically pays directly.

should you take it?#

If you will ever fit models to real data — biostatistics, insurance, econometrics, machine learning — this is core material wearing an elective’s course code. GLMs are the lingua franca of applied statistics, and this course teaches the machinery honestly instead of hiding it behind glm(). The prerequisite load is light, but you want to be comfortable with likelihood-based inference and basic linear algebra before term starts.

flashcards#

the 180-card MATH5806 Anki deck built alongside these notes — exponential family, estimation and inference, GLM practice with worked R output cards — is for sale on the bazaar (A$5.99, .apkg).