Getting A Step Ahead: Using the Regularized Horseshoe Prior to Select Cross-Loadings in Bayesian Regularized Structural Equation Modeling (SEM)

This repository contains the code of my Master's Thesis in Methodology & Statistics at Utrecht University (September '21 - June '22).

You can clone the repository by running:

git clone https://github.com/JMBKoch/1vs2StepBayesianRegSEM/.

• The most recent version of the research article reporting the results of this project can be found on /Rmd/thesis.

• Note that all scripts assume that this repository has been cloned to the home directory of a unix-based system. Hence, if you're on Windows or you want to work from a different path, you will have to adjust the paths in R/main.R and in Rmd/analyses manually.

• The simulation study can be conducted by sourcing or running R/main.R. Note that all study-parameters, including the MCMC sampling parameters, and the number of clusters used in the parallelization are specified in R/parameters.R. R/functions.R contains all functions that are used in R/main.R. If you want to re-run the simulation, please first uncomment line 28 & 29 in R/main.R. This ensures that the output is removed and newly saved. Otherwise the new results will be appended to the old ones.

• Packages should be installed automatically, if they are not yet. However, this may not work on all systems/ versions of R. Hence, if the script does not run checking if the packages are installed correctly may be a sensible first step in the debugging process. An overview of the required packages can be found at the top (line 7-13) of R/parameters.R.

• In order for cmdstanr to work, it is required to run cmdstanr::install_cmdstan() a single time.

• Note that if the model is adjusted, the code in stan needs to be adjusted accordingly as well.

• data contains the raw datasets that were simulated based on the population conditions. It will be simulated and saved again when running R/main.R.

(c) J.M.B. Koch, 2022 (updated 2024)