lvmcomp2

lvmcomp2 is an R package for fast Estimation for Latent Variable Models using the Unified Stochastic Proximal (USP) Algorithms.

It provides stochastic approximation algorithms for latent variable models with a high-dimensional latent space and with constraints/penalized parameters. So far, we provide functions for confirmatory item factor analysis based on the multidimensional two parameter logistic (M2PL) model. The computation is facilitated by multiprocessing 'OpenMP’ API. For more information, please refer to:

Zhang, S., & Chen, Y. (2022). Computation for Latent Variable Model Estimation: A Unified Stochastic Proximal Framework. Psychometrika.

Major Features

Fast. By using state-of-the-art proximal quasi-Newton SA algorithms, lvmcomp2 is much quicker than commonly implemented numerical integration methods or the Stochastic EM algorithm.
Generic. Include a general class of algorithms capable of handling many latent variable models and can be easily extended to more general cases.
Written concisely in a combination of R and C.

Download

lvmcomp2 package can be installed simply by running the following command in R console.

> remotes::install_github(“slzhang-fd/lvmcomp2”)

Source code can be found on lvmcomp2's GitHub page.

Example

A toy example for estimating the confirmatory IRT model using the proposed USP method.