Distinguishing between latent classes and continuous factors: Resolution by maximum likelihood? (original) (raw)
Related papers
Special issue on Advances in latent variables: methods, models and applications
Advances in Data Analysis and Classification, 2016
Starting from Spearman's 1904 pioneering work on factor analysis, latent variable models have witnessed an ever increasing, even though sometimes controversial, diffusion in the statistical literature. They have been extended to deal with different kinds of data structures, and thereby helped to analyse more and more complex situations. Finally, they turned out to be both a powerful instrument for a better understanding of reality and a necessary tool to perform dimension reduction. With the development of refined latent variable models new computational algorithms have been designed that rendered the corresponding parameter estimation fast and reliable. New research lines have incorporated latent variables as a necessary building block. A class of latent variable models that has been deeply studied and widely applied is Latent Class Analysis, which aims at modelling and testing the existence of latent subgroups based on the association among a set of discrete observed variables. The key assumption is conditional independence of the observed variables given the latent classes. This strong assumption largely simplifies the model and prevents it to be affected by the curse of dimensionality, but as for any strong assumption, its violation may lead to poor model performances and to inconsistent results. Although widely used and explored the research area related to Latent Class Models still leaves space for advances and new solutions in terms of modelling strategies, ability to deal with complex data structures (e.g. multilevel or longitudinal) or with mixed type variables, and development of new methods to test model fit. This Special Issue of ADAC, entitled Latent Variables: Methods, Models and Applications has been designed to collect a range of innovative and high quality research B Maurizio Vichi
Second special issue on “Advances in latent variables: methods, models and applications”
Advances in Data Analysis and Classification, 2016
It has been designed to collect a range of innovative and high quality research papers on new challenges and recent developments in the field of latent variable models and their application to real problems. The first Special Issue on this topic, i.e., Issue 2 of vol. 10 (2016) of ADAC presented papers on multilevel latent class models, mixture models for mixed-type data in model-based clustering, latent class CUB models, latent class growth models, model selection, testing and item selection in Latent Class Models. The term latent-from the present participle of the Latin verb "latere" that means "being hidden"-denotes a variable that underlies a phenomenon (hypothetical construct) which cannot be directly observed and therefore must be inferred (reconstructed) from other variables that can be directly observed (manifest variables), by means of statistical methods and models. Latent variable methods and models can be classified according to whether the manifest and latent variables are categorical or continuous. In the case of continuous latent variables, Factor Analysis is based on a formal model for predicting observed variables from latent factors and has been frequently combined, during the last twenty years, with cluster analysis techniques in order to obtain a simultaneous reduction in the number of both objects and variables. Along such lines, the first two papers of the present ADAC issue deal with factor analysis
Journal of Pediatric Psychology, 2014
Objective Pediatric psychologists are often interested in finding patterns in heterogeneous cross-sectional data. Latent variable mixture modeling is an emerging person-centered statistical approach that models heterogeneity by classifying individuals into unobserved groupings (latent classes) with similar (more homogenous) patterns. The purpose of this article is to offer a nontechnical introduction to cross-sectional mixture modeling. Method An overview of latent variable mixture modeling is provided and 2 cross-sectional examples are reviewed and distinguished. Results
Frontiers in psychology, 2014
The purpose of this study was to examine in which way adding more indicators or a covariate influences the performance of latent class analysis (LCA). We varied the sample size (100 ≤ N ≤ 2000), number, and quality of binary indicators (between 4 and 12 indicators with conditional response probabilities of [0.3, 0.7], [0.2, 0.8], or [0.1, 0.9]), and the strength of covariate effects (zero, small, medium, large) in a Monte Carlo simulation study of 2- and 3-class models. The results suggested that in general, a larger sample size, more indicators, a higher quality of indicators, and a larger covariate effect lead to more converged and proper replications, as well as fewer boundary parameter estimates and less parameter bias. Furthermore, interactions among these study factors demonstrated how using more or higher quality indicators, as well as larger covariate effect size, could sometimes compensate for small sample size. Including a covariate appeared to be generally beneficial, alt...
Principles and Applications of Latent Class Analysis in Psychological Research
The Irish Journal of Psychology, 2007
This paper provides an overview of latent class analysis (LCA) and its utility in contemporary psychological research. Rather than focussing on the statistical and mathematical underpinnings of this method of analysis, a nontechnical presentation of conceptual issues is offered. The paper therefore and why a psychologist may want to use LCA. A worked example of how some examples from the research literature are presented.
A REVIEW ON APPLICATIONS OF LATENT CLASS ANALYSIS
Latent class analysis (LCA) is considered to be an equivalent methodology for Factor Analysis, typically used for dichotomous or polytomous variables. The parameters of interest in a typical problem of latent class analysis are the unobserved proportion or size of the latent classes and the conditional item-response probabilities given the membership in a latent class. Based on the observed data on manifest variables, LCA provides a classification among the population. Manifest variables are called the "indicators" of a particular latent class. The present paper provides a review on the theory of latent class analysis and its wide area of applications in various disciplines such as
How To Use Latent Analyses Within Survey Data Can Be Valuable Additions to Any Regression Model
2014
The current study looks at several ways to investigate latent variables in longitudinal surveys and their use in logistic regression models. Three different analyses for latent variable discovery will be briefly reviewed and explored. The procedures explored in this paper are PROC LCA, PROC LTA, PROC CATMOD, PROC FACTOR, PROC TRAJ, and PROC SURVEYLOGISTIC. The analyses defined through these procedures are latent profile analyses, latent class analyses, and latent transition analyses. The latent variables will then be included in three separate regression models. The effect of the latent variables on the fit and use of the logistic regression model compared to a similar model using observed data will be reviewed. The data used for this study was obtained via the National Longitudinal Study of Adolescent Health, a study distributed and collected by Add Health. Data was analyzed using SAS 9.3. This paper is intended for any level of SAS user. This paper is also written to an audience w...
Two Studies of Specification Error in Models for Categorical Latent Variables
2016
This article examines the problem of specification error in 2 models for categorical latent variables; the latent class model and the latent Markov model. Specification error in the latent class model focuses on the impact of incorrectly specifying the number of latent classes of the categorical latent variable on measures of model adequacy as well as sample reallocation to latent classes. The results show that the clarity of remaining latent classes, as measured by the entropy statistic depends on the number of observations in the omitted latent class—but this statistic is not reliable. Specification error in the latent Markov model focuses on the transition probabilities when a longitudinal Guttman process is incorrectly specified. The findings show that specifying a longitudinal Guttman process that is not true in the population impacts other transition probabilities through the covariance matrix of the logit parameters used to calculate those probabilities.