RLS is an R package designed to compute reweighted least squares (RLS) test statistics and their p-values. The package is currently hosted in my GitHub repository. To download and install the RLS package, simply follow these steps: install.packages("githubinstall") library(githubinstall) install_github("bqzheng/RLS", dependencies = TRUE) library(RLS) RLS is crafted to address the issue of over-rejection by maximum likelihood (ML) estimator and under-rejection by generalized least squares (GLS) in structural equation modeling (SEM) and covariance structure analysis (CSA), including confirmatory factor analysis (CFA). RLS delivers highly consistent test statistics, proving particularly advantageous in situations with small sample sizes in normal data. The robust versions of RLS also outperform those of ML and GLS in handling non-normal data and misspecified analysis models. For insights into the statistical properties and applications of RLS, please consult the following research:
Regularized Generalized Least Squares (RGLS) Estimator
RGLS is an R package designed to compute Regularized Generalized Least Squares test statistics and their corresponding p-values. This package is hosted in my GitHub repository. To install it, use this command install_github("bqzheng/RGLS", dependencies = TRUE). The RGLS function performs a non-linear transformation on sample eigenvalues, providing a robust method for stabilizing covariance estimation. The key concept behind covariance estimation regularization involves extracting eigenvalues from a potentially ill-conditioned or singular covariance matrix and applying regularization through a quadratic function. This regularization process pushes down the highest eigenvalues and pulls up the lowest eigenvalues. For details about the statistical properties and applications of RGLS, please consult the following research: