LCA News

LCA Stata Plugin

April 2, 2014

The Methodology Center is pleased to release the latest version (1.1) of the LCA Stata plugin for conducting latent class analysis (LCA). The software is available for download free of charge. For an overview of the functionality of the LCA Stata plugin, please visit the download page. The new version includes functionality requested in our recent software survey, including

  • the ability to assess identification of models with covariates via multiple random starts,
  • an indication of which latent class is the best match for each individual, and
  • the option to generate 20 random draws for each individual’s class membership based on posterior probabilities.

The users’ guide has also been updated and revised based on user feedback. Please email mchelpdesk@psu.edu with any questions.

 

Read more or download the software

PROC LCAJanuary 24, 2014

We have released PROC LCA v. 1.3.1 to fix a bug related to the seed_draws statement in v. 1.3.0 that sometimes causes an error. If you used seed_draws and did not receive an error message, then the output generated is reliable. There is no change in functionality between v. 1.3.0 and 1.3.1.

Download PROC LCA v. 1.3.1

Bethany BrayStephanie LanzaSeptember 24, 2013

Latent class analysis (LCA) has become a popular tool for identifying subgroups within populations. Despite the fact that these are latent subgroups and, therefore, membership in each class is unknown, research questions sometimes make it necessary to assign individuals to classes and then treat class membership as known in later analyses. Current standard practice is to retain from LCA each individual’s posterior probabilities of class membership, and then either assign them to the most likely class or to repeatedly assign them using pseudo-class draws. These practices, however, are known to attenuate estimates in subsequent analyses.

 

In a new article in Structural Equation Modeling, "Eliminating bias in classify-analyze approaches for latent class analysis," by Center Investigators Bethany Bray and Stephanie Lanza and Xianming Tan of McGill University, the authors propose a straightforward solution that includes all variables used in the subsequent analyses as covariates in the LCA. Then, class assignment is carried out using either the most likely class or pseudo-class draws. With adequate measurement quality and sample size, this approach substantially reduces the bias that results from common approaches to class assignment.

Nicole ButeraStephanie LanzaDonna Coffman

August 1, 2013

Donna Coffman and Stephanie Lanza lead the Methodology Center’s research projects on causal inference and latent class analysis (LCA), respectively. Recently, they began integrating modern causal inference techniques with LCA to reveal the impact that nonrandomized risks have on later, complex, behavioral outcomes. In a new article in Prevention Science, “A Framework for Estimating Causal Effects in Latent Class Analysis: Is There a Causal Link Between Early Sex and Subsequent Profiles of Delinquency?” Center researchers Nicole Butera, Stephanie, and Donna examine data on 1,890 adolescents from the National Longitudinal Study of Adolescent Health to determine whether early sexual initiation impacts later profiles of delinquent behavior.

Linda Collins, Stephanie Lanza, Donna Coffman, Bethany Bray, Kari Kugler, and Anne Fairlie

Society for Prevention Research (SPR) Annual Meeting

San Francisco, CA, May 28-31, 2013

 

The Methodology Center will be active at the upcoming annual meeting of the Society for Prevention Research (SPR).  At this year's conference, The Science of Prevention: Building a Comprehensive National Strategy for Well-Being, we will be presenting symposiums, a special interest group, paper talks, and multiple posters. Also look for us at the ECPN symposiums. We hope to see you there!

April 16, 2013

PROC LCA & PROC LTAThe Methodology Center is pleased to release the latest version (1.3.0) of the SAS procedure PROC LCA for conducting latent class analysis (LCA). The software is available for download free of charge. The download also contains PROC LTA, the SAS Procedure for latent transition analysis. For an overview of PROC LCA and PROC LTA features, please visit the PROC LCA download page. New PROC LCA features include

  • Ability to assess identification of models with covariates via multiple random starts
  • Indication of which latent class is the best match for each individual
  • Option to generate 20 random draws for each individual’s class membership based on posterior probabilities

The users’ guide has also been updated and revised. Please email mchelpdesk@psu.edu with any questions.

Download the software or read more

Power Curves for LCAMarch 28, 2013

Latent class analysis (LCA) is a tool used by behavioral scientists to identify subgroups within a population. When conducting LCA, choosing the number of classes (subgroups) in the model is a critical step. Some software, like the Methodology Center’s Bootstrapping Macro for PROC LCA, allows users to perform the bootstrap likelihood ratio test (BLRT) to evaluate models with different numbers of classes. Until now, however, there has been no way to determine the sample size needed to provide adequate power for the BLRT. In a new article by  John Dziak, Stephanie Lanza, and Xianming Tan to appear in Structural Equation Modeling, the authors provide effect-size measures and power curves that can be used to predict power for the BLRT in LCA.  These power curves can guide researchers in determining the sample size needed for their proposed LCAs.

College enrollment does not lead to problem drinking in adulthood March 25, 2013

Despite high levels of binge drinking that take place on college campuses, recent evidence suggests that college enrollment does not lead to substance abuse problems later in adulthood, and it may actually prevent adult substance abuse among youth who would not be expected to attend college, according to Methodology Center Investigators Stephanie Lanza and Donna Coffman. "College is often perceived as a risky environment for problem drinking, but seldom have people looked at the long-term consequences of attending college on substance-use patterns," Stephanie said.

 

This research is described in the article, “Causal Inference in Latent Class Analysis” which will appear in a forthcoming issue of Structural Equation Modeling. As the title indicates, the paper’s primary focus is methodological; this is the first time propensity score methods have been used to conduct causal inference in latent class analysis. This approach could be used to answer a large number of questions about possible determinants of complex behaviors.

Read more

LCA Stata PluginFebruary 25, 2013

The Methodology Center is pleased to announce our first software release compatible with Stata, the LCA Stata Plugin for conducting latent class analysis (LCA). The software is available for download free of charge from our software page. The LCA Stata Plugin is based on version 1.2.7 of PROC LCA (also developed by the Methodology Center) and includes the same functionality. Features include

  • multiple-groups LCA
  • LCA with covariates (prediction of latent class membership)
  • posterior probabilities available in output
  •  sampling weights and clusters can be incorporated

 

The software now has a full users’ guide. Please email mchelpdesk@psu.edu with any questions.

 

Download the software or read more

Stephanie Lanza & Bethany BrayFebruary 21, 2013

We are pleased to announce that Methodology Center Investigators Stephanie Lanza and Bethany Bray will present this year’s Summer Institute on Innovative Methods, “Introduction to Latent Class Analysis.”

 

Sponsored by the Methodology Center and the National Institute on Drug Abuse, the 18th Summer Institute will present the theoretical background and applied skills needed to address interesting research questions using latent class analysis (LCA). By the end of the workshop, participants will have fit preliminary latent class models to their own data. The institute will be held on June 27-28, 2013 at Penn State in State College, PA. Apply early; there are a very limited number of seats available.

 

Read more or apply.

Follow Us