Jacques A. Hagenaars and Allan L. McCutcheon, Eds. Applied Latent Class Analysis

Cambridge University Press, 2002, 454 pp. Latent class models can provide a useful summary of data for which an observed set of categorical variables are highly related and represent some underlying concept. In their basic form these models are the categorical data analogues to factor analysis and structural equation modeling with latent variables. The possible applications of such models in the social sciences are numerous. One example would be using a set of categorically scored questionnaire items to group respondents according to various personality types. In this case, we would assume that the underlying personality type has caused responses to the various questions. Although latent class models were first used several decades ago, they are still not commonly applied, perhaps largely because they are not widely understood. Applied Latent Class Analysis, an edited volume that includes contributions from some of the leading researchers in the field, could help fill some of this gap in knowledge. Applied Latent Class Analysis is an outstanding book that demonstrates the potential uses of latent class models. Just as importantly, it provides insight into some recent innovations with respect to these models. The editors of the book, Jacques Hagenaars (Tilburg University) and Allan McCutcheon (Gallup Research Center, University of Nebraska, Lincoln), are experts in the field who have assembled a truly first-rate set of articles. Reflecting the major uses of these models, the book is divided into four sections: (1) introduction, (2) classification and measurement, (3) causal analysis and dynamic models, and (4) unobserved heterogeneity and nonresponse. Each of these sections contains insightful chapters. The first section of the book contains chapters by McCutcheon and Leo Goodman. Goodman's chapter gives a good overview of latent variable models. Using examples that will be familiar to most social scientists, it provides an interesting history of the models and how they are related to similar models. McCutcheon's chapter is a very good introduction to the standard latent class model. Both of these chapters should prove helpful to a newcomer to these models. The second section discusses more specifically how latent variables can be seen as unobserved constructs represented by observed variables. Various ways of specifying the underlying variables--whether they are nominal, ordered or interval level variables--are discussed. We learn from Vermunt and Magdison in chapter three that models assuming a nominal latent variable have a strong similarity to some cluster analysis techniques. Chapter five, by Marcel Croon, is particularly insightful in its discussion of ordered latent class models as they can be applied to rating scales. The third section on causal analysis is perhaps the most useful. In chapter eight, Dayton and Macready describe blocking models, which are effective in determining the influence of continuous and categorical variables on the probability of observations belonging to particular latent classes. These models are simply adaptations of the standard logit model. Hagenaars extends this idea further to show how loglinear or logit models can be used to handle systematic measurement errors. These models parallel structural equation models with latent continuous variables. …

Fetched live from OpenAlex and de-inverted. Abstracts are not stored in this database: the inverted indexes are 8.6 GB of the frame’s 9.3 GB of text, and the host has 13 GB free.