Individuals differ in many substantive ways that are not always captured through the assumed data-generating model. In an effort to address this, we present a robust estimation procedure based on the EM algorithm which we call REM (robust expectation-maximization).

This folder contains R package files to run exploratory and confirmatory factor analyses (EFAs and CFAs) using the REM algorithm. The following code can be used to download the latest version of the package to your RStudio from Github.

library(devtools)
devtools::install_github('knieser/REM/R_package')
library(REMLA)

REMLA tutorial

This folder contains the MATLAB code used for simulation studies found in the paper, Addressing Heterogeneous Populations in Latent Variable Settings through Robust Estimation. Further details about REM and the simulation studies can be found there.

In MATLAB, with an input p-by-n dataset X, REM estimation for

• Gaussian mixture models can be run from Psych_Methods_paper/MixtureModel/GMM_estimates.m
• Exploratory factor analysis can be run from Psych_Methods_paper/FactorAnalysis/FA_estimates.m

Simulations can be run from Psych_Methods_paper/MixtureModel/GMM_sim_main.m and Psych_Methods_paper/MixtureModel/FA_sim_main.m for mixture models and factor models, respectively

The code in xxprelimR_code_for_mixture_models/ folder adds functions to run REM estimation with Gaussian mixture models and is currently under development.