Latent Variable Module

Latent variable models have been popularized at Hopkins by Dr Karen Bandeen-Roche, Associate Professor in the Departement of Biostatistics. She has made substantial contributions in this area and has provoked students' interest. Numerous latent variable programs have been written by several different authors, many of whom were/are Dr Bandeen-Roche's students. Originally, Dr Bandeen-Roche wrote Gauss programs to perform latent class analysis and Diana Miglioretti wrote SAS programs to perform latent class regression. These programs have since been translated into Splus and now include some error checking procedures. Both Liz Garrett and Qian-li Xue have written Splus programs to perform latent class analysis and regression. Dr Bill Rising and Brent Johnson have since took some of the current LCA programs, improved on the computing efficiency and added some error checking procedures to create a so-called "user-friendly" LCA function. Currently, we are aiming to create a single latent variable class to handle a wide variety of data, models, and problems.

Latent Class Analysis

These functions perform standard (unrestricted) latent class analysis on binary data. They take an N x M matrix of 0/1 responses and a hypothesized number of classes, J, as arguments and return a vector of prevalences for each of the J classes and an M x J matrix of posterior probabilities.

Splus programs (for UNIX)

You must unzip (gzip -d filename.gz) and untar (tar xvf filename.tar) this file.

Splus programs (for UNIX or PC)

Download all four files: gop.S , lca.S , isallequal.S , post.prob.bin.S .

SAS programs

Latent Class Regression

Coming Soon!