• HOME
• Research
• Teaching
• Software
• Personal
• SMfSB

• Lecture notes
• MATH4341
• APTS StatComp

Teaching / APTS StatComp

This is the webpage supporting the 2022/23 iteration of the APTS course on Statistical Computing, which took place in Warwick, December 2022. This was my final time delivering the course. I will leave the materials here for the convenience of the five cohorts of PhD students who took my version of this course (and anyone else who is interested), but note that the person taking over this course is likely to update it, so please don't assume that future iterations will be very similar.

The APTS courses are aimed at first year UK PhD students, in statistics and related fields.

• APTS Website
• Statistical Computing
• Module resources
• Preliminary material: PDF or accessible HTML
• Lecture notes: PDF or accessible HTML, and R code extracted from the notes
• Computer Lab 1 - Wednesday
• Computer Lab 2 - Friday
• Slides for bonus lecture (if time permits): On future directions for statistical computing. The full recorded video for this bonus lecture is available via the link below. But it is quite long, so if you would prefer a shorter, more Bayesian, and more up-to-date version of this talk, the slides and video of my recent Laplace's Demon talk are also available.
• Lecture videos (recorded for 2020)
• Assessment - distributed by the APTS organisers

Some links and further reading

Note that there are many links and suggested sources of further reading in the main course notes. Below I just include a few especially relevant sources, or useful sources omitted from the notes.

• My lecture notes on multivariate data analysis covers similar ground to Chapter 2 of this course (Matrix computation), but in a bit more detail, and with a bit more discussion of matrix decompositions and especially their applications to statistical computing.
• The Matrix Cookbook is an invaluable source of matrix identities, etc. The results are especially useful for figuring out non-trivial gradients of scalar functions.
• Regarding Chapter 5 (Random number generation), relatively little was said about the generation of non-uniform random variates. In this context, Luc Devroye's book on Non-Uniform Random Variate Generation is a classic reference.
• Nick Higham's blog - an excellent source for scientific computing, numerical linear algebra, and numerical analysis. See especially his what is ... series, and his seven sins of numerical linear algebra.
• Herbie can help with re-writing floating point expressions to make them more numerically stable
• For more on functional programming for statistical computing, see my blog and especially the series on Bayesian inference for a logistic regression model (with accompanying repo).