PCLVMthesis implements Partially Confirmatory Latent Variable Modelling (PCLVM) developed for Zhang Yifan’s PhD thesis:

Bayesian Regularization for Latent Variable Modeling: A Unified Framework and Prior Comparison

The methodological foundation builds upon the pcfa function from the LAWBL package. The current implementation re-engineers the estimation procedure using C++ improved computational efficiency and scalability, while introducing substantial extensions in model structure and estimation strategy. The R6-based interface provides a user-friendly front-end, while the underlying Rcpp module performs posterior sampling and matrix operations efficiently, ensuring extensibility and computational robustness.

Key Features

  • Applies regularization to all major components of the factor model, including factor loadings, latent factor covariance matrices, and residual covariance matrices, with graphical regularization used for covariance structures when appropriate.

  • Supports partially confirmatory loading specifications through a user-defined design matrix Q, enabling structured theoretical constraints alongside regularized estimation.

  • Provides flexible latent factor structure modeling through full or block-specific Inverse-Wishart priors, full or block-specific covariance shrinkage, and partial regularization on selected correlation blocks combined with Inverse-Wishart estimation.

  • Offers multiple shrinkage prior choices for each model component, including Lasso, Horseshoe, and Spike-and-Slab (SSP), with independent user control over prior specification across parameter blocks.

  • Accommodates multiple latent variable structures, including standard CFA, bifactor models, MTMM models, and testlet effect models within a unified partially confirmatory framework.

  • Supports ordinal indicators, mixed measurement scales, and missing data handling through Bayesian data augmentation.

These capabilities allow structured theoretical specifications and data-driven shrinkage to coexist within a unified Bayesian framework for latent variable modeling.

Installation

# install.packages("devtools")
devtools::install_github("ZhangYifan-Jenny/PCLVMthesis")

Basic Usage

Please refer to the online tutorial for more comprehensive examples.

References

  • Zhang, Y. (PhD Thesis). Bayesian Regularization for Latent Variable Modeling: A Unified Framework and Prior Comparison.

  • Zhang, Y., & Chen, J. (2024). Accommodating and Extending Various Models for Special Effects Within the Generalized Partially Confirmatory Factor Analysis Framework. Applied Psychological Measurement, 48(4-5), 208-229.

  • Chen, J. (2020). A partially confirmatory approach to the multidimensional item response theory with the Bayesian Lasso. Psychometrika. 85(3), 738-774. DOI: 10.1007/s11336-020-09724-3.

  • LAWBL Package: https://jinsong-chen.github.io/LAWBL/index.html

Development Status

This package is under active development and intended for research use.