Latent Class What? the What, When, Why, and How of Latent Class Modeling with Categorical Data Using R

Social work has a longstanding interest in understanding human behavior in general, and client outcomes in particular, as a function of the person, environment, and interaction between person and environment through person-in-environment perspective (Kondrat, 2002). However, social work researchers and those in related areas have struggled with analyzing data from studies in which individuals must be classified to understand class-related outcomes. Latent Class Analysis (LCA) can assist researchers with the classification of their data using a categorical latent variable approach. These models allow investigators to analyze multivariate categorical data and test meaningful hypotheses regarding participant classification and the variables that predict class membership. This workshop will build upon the well-received series on latent variable modeling that the authors have offered over the last four years at SSWR.

We will begin by introducing LCA and relevant mixture modeling techniques, what they are conceptually, and why social work researchers can use them. In doing this we will discuss the rationale for LCA, circumstances under which it is useful, and the types of research questions they can answer. We will use a publically available cross-sectional dataset to illustrate basic LCA concepts (e.g., assumptions, model building, data management, data and formula definition, and estimation), and we will show how much of what participants know about GLM and GZLM techniques is applicable. Once we demonstrate model estimation, we will build on this by explaining classification regression, a form of multinomial logistic regression used to identify predictors of class membership.

We will then provide an introduction to LCA using R, a free, open-sourced language for statistical analyses and graphical production. R provides a number of advantages that likely have benefits for social work practitioners and researchers, such as its packages and their adaptable resources, its flexibility when executing statistical functions, its ability to simply integrate statistical output in publishable documents (e.g. LaTex and Publisher), and its easy-to-use graphical interface (Dalzell, 2013; Muncheon, 2009). As a free tool available for statistical analyses, social workers can easily access R and use it to perform multilevel analyses. This workshop will teach participants the basics of the R language, such as data entry and manipulation commands, simple graphical procedures, and model identification. We will then present a “walk-through” of how to perform multilevel analyses using the poLCA (Polytomous Variable Latent Class Analysis) (Linzer, n.d.), introducing helpful data manipulation functions for common LCA applications, model estimation, generating illustrations, and interpreting and reporting results. We will conclude by providing a number of additional resources (i.e. R scripts, package information and applicable datasets, collected R literature).

This workshop will provide an introduction to a freely accessible and powerful statistical tool that will allow social workers to incorporate latent class analysis and related techniques into their research and evaluation efforts. As we have done with our other workshops, we will give participants materials from our latent variable modeling course, which will include a bibliography of relevant texts and articles, PowerPoint presentations, datasets, and helpful resources, for those who want to learn more about LCA.