SISG: Module S2

Module S2: Bayesian Statistics for Genetics

Instructors: Ken Rice and Steve Qin

This page, updated throughout the course, will feature slides from our sessions, and examples for you to try.

Prior to the module, please install up-to-date versions of R and RStudio on the laptop you will use during the summer institute. Both are free.

R is available from this list of sites.
RStudio is available here.

Other software that will be useful, but only used in parts of the module:

Stan: Windows users first need the Rtools toolchain, Mac users may need the toolchain and corresponding tools. After ensuring you have up-to-date versions of R and Rtools, install the rstan package - but direct from the Stan project, not CRAN. To do this use the 'Getting started' instructions here
JAGS: Download and install JAGS itself , then install R's rjags package
INLA: To get the INLA package for R, follow the non-standard procedure here; INLA may also need earlier versions of R
WinBUGS: you will need WinBUGS and the R2WinBUGS R package. Note making WinBUGS work on a Mac requires a package manager such as HomeBrew, and you may prefer just to use JAGS instead.

Slides and exercises

Script files will posted following each session; these contain our R code for the exercises. To make them work on your computer, remember to modify file names and locations appropriately. Also note that many different 'correct' solutions are possible.

The module has 10 sessions, each of 90 minutes. The standard format for a session is approximately:

50 minutes of lecture material (presented live, also recorded for later reference)
25 minutes of code demonstrations and exercises for you to try. We strongly recommend working with other students
15 minute discussion of exercises, where the instructors will present possible solutions and answer questions (also recorded)

Slides and code used in-class

Wednesday, June 4th
Time	Topic	Lecture	Exercises/Discussion
1:00pm-3:00pm	0/1. Introductions, motivation for Bayes	Slides [.pdf] Slides [.pdf]	Exercises, Key [.pdf]
3:30pm-5:00pm	2. Binomial sampling, part 1	Slides [.pdf]	Demo and exercises [.Rmd], [.pdf], Key [.pdf]
Thursday, June 5th
Time	Topic	Lecture	Exercises/Discussion
8:30am-10:00am	3. Binomial sampling, part 2	Slides [.pdf]	Demo and exercises [.Rmd] [.pdf]
10:30am-12:00pm	4. Multinomial Samples & INLA	Slides [.pdf], INLA example [.R]	Demo and exercises [.Rmd] [.pdf]
1:30pm-3:00pm	5. Linear models	Slides [.pdf]	Exercises Script from demo [.R]
3:30pm-5:00pm	6. Clustering	Slides [.pdf]	No exercises for this session
Friday, June 6th
Time	Topic	Lecture	Exercises/Discussion
8:30am-10:00am	7. Meta-analysis	Slides [.pdf]	Exercises Script file [.R]
10:30am-12:00pm	8. Hierarchical models	Slides [.pdf]	No exercises for this session
1:30pm-3:00pm	9. Testing and Multiple Testing	Slides [.pdf], prostmat.csv, prostz.txt	Demo [.Rmd] (no exercises for this session),
3:30pm-5:00pm	10. Software examples	Slides [.pdf]	See demo resources below

Resources for Session 10: most of the examples run the HWE example - each example is self-contained

WinBUGS example - requires installing WinBUGS.
R2WinBUGS and JAGS example - requires installing R and its R2WinBUGS package. For the JAGS part you will also need to install JAGS itself and the rjags package
Stan example - requires installing R. Windows users will also need Rtools, Mac users may need the toolchain and corresponding tools
INLA examples - requires installing R. To get the INLA package for R, follow the non-standard procedure here
Examples with DIC/WAIC, using the FMS data [.txt] from Andrea Foulkes' book

Datasets

ASAgene.txt: Data for Session 2 and 3
diagen_bma: Data and posterior simulations for Session 6 - load into an R session with load()
yX_FTO: Copy of the FTO data, for Session 6 - load into an R session with load()

Other resources

The R site, which includes the Comprehensive R Archive Network (CRAN) of downloads and packages
A reference card of common R commands (and a slightly longer reference card)

Some recommended books;

A First Course in Bayesian Statistical Methods (Hoff, 2009)
Bayesian and Frequentist Regression Methods (Wakefield, 2013) - available from with UW login
Foulkes, Applied Statistical Genetics with R For Population-based Association Studies, 2009; this book covers a lot of the material we'll discuss, although seldom from a Bayesian point of view.