Bayesian dimensionality reduction with PCA using penalized semi-integrated likelihood

Overview

In this repository you can find a code that can be used to recreate figures from the paper Bayesian dimensionality reduction with PCA using penalized semi-integrated likelihood.

Important informations

1. Running this code requires both R and MATLAB. In files figure_2.R, figure_3.R, data_generation_robustness_fixed_effects.R one needs to specify path to MATLAB executable file.
2. Please note that it might take a lot of time to run simulations (several days). We recomend to separate performing simulations and plotting results. To speeed up calculations one might reduce number of repetitions (numb.repetitions), reduce the set of number of variables included in simulations (vars) or reduce the set of tested signal to noise ratios (SNRs).
3. Packages required for this simulations:
   • pesel
   • FactoMineR
   • pryr
   • softImpute
   • ggplot2 (for plots)

To install all those dependencies just run the code below in R console

library(devtools)
install_github("psobczyk/pesel")
install.packages(c("pryr", "softImpute", "FactoMineR", "ggplot2"))