by Daniel Runcie, Sayan Mukhergee

Reference: Runcie, D., & Mukherjee, S. (2013). Dissecting High-Dimensional Phenotypes with Bayesian Sparse Factor Analysis of Genetic Covariance Matrices. Genetics, 194(3), 753–767. http://doi.org/10.1534/genetics.113.151217

Published version in MATLAB on website:

• includes Ayroles_et_al_Competitive_fitness, Simulations with half-sib design
• should be able to replicate all analyses from paper (up to Monte-carlo error in Gibbs and in simulations)

• Fixed calculation of genetic and interaction specific effects. The calculation and corresponding text of the paper missed $A^{-1}$. This mistake, as well as other errors in the paper are documented here. A re-analysis of the simulations presented in the paper and an updated Appendix are presented here.

Nearly complete re-write of the model code, but should maintain identical function \

• variables have been re-named to more closely correspond to the paper
• sampler function has been re-written to only sample.
• A new function initializes the sampler, only run once
• sampler function starts where the previous run left off (including maintaining the random number generator), so should be the same as running one continuous chain

R clone of V2.0 Matlab code

• Functionality should be identical. Worth checking. Note that the RNG is different.
• embeded in the Gibbs sampler are two versions of each sampler function, a native R version, and a Rcpp version. They should be identical (up to RNG differences). The Rcpp function has the same name and arguments, but with "_c" appended to the function name.