lcca-package {lcca} | R Documentation |
Functions for fitting latent-class models to polytomous items with or without covariates, and for estimating average treatment effects in nonrandomized studies when the treatment is a latent class.
This package provides functions for latent-class analysis (LCA) with and without covariates. It differs from other available LCA implementations (notably, the R package poLCA) in the following respects.
Computations are performed in native Fortran.
Standard errors for parameters are computed by several different methods.
The functions for latent-class modeling support multi-group analyses, which are useful for examining questions of measurement invariance.
The estimation procedures accept survey weights, computing pseudo-maximum likelihood (PML) estimates for model parameters.
Standard errors are computed for the popular class of with-replacement (WR) designs, using a linearization (sandwich) method.
We also provide functions for latent-class causal analysis (LCCA), implementing new procedures for estimating average treatment effects when the treatment is a latent class. The major functions are:
lca Latent-class analysis lcacov Latent-class analysis with covariates lcca Latent-class causal analysisThe package also includes two datasets:
hivtest Diagnostic tests for HIV infection abortion Abortion attitudes from the General Social Survey NHsmoking Recent cigarette use from NHANES diet Simulated dieting study
Joseph L. Schafer
Maintainer: The Methodology Center <mchelpdesk@psu.edu>