Predicting Latent Class Scores for Subsequent Analysis

Janne Petersen, Karen J Bandeen Roche, Esben Budtz-Jørgensen, Klaus Groes Larsen

Research output: Contribution to journalArticle

Abstract

Latent class regression models relate covariates and latent constructs such as psychiatric disorders. Though full maximum likelihood estimation is available, estimation is often in three steps: (i) a latent class model is fitted without covariates; (ii) latent class scores are predicted; and (iii) the scores are regressed on covariates. We propose a new method for predicting class scores that, in contrast to posterior probability-based methods, yields consistent estimators of the parameters in the third step. Additionally, in simulation studies the new methodology exhibited only a minor loss of efficiency. Finally, the new and the posterior probability-based methods are compared in an analysis of mobility/exercise.

Original languageEnglish (US)
Pages (from-to)244-262
Number of pages19
JournalPsychometrika
Volume77
Issue number2
DOIs
StatePublished - Apr 2012

Fingerprint

Latent Class
Covariates
Latent Class Model
Posterior Probability
Maximum likelihood estimation
Consistent Estimator
Maximum Likelihood Estimation
Exercise
Psychiatry
Disorder
Minor
Regression Model
Simulation Study
Methodology

Keywords

  • classification
  • latent class regression
  • latent class scores
  • least squares class
  • three-step procedure

ASJC Scopus subject areas

  • Psychology(all)
  • Applied Mathematics

Cite this

Predicting Latent Class Scores for Subsequent Analysis. / Petersen, Janne; Bandeen Roche, Karen J; Budtz-Jørgensen, Esben; Larsen, Klaus Groes.

In: Psychometrika, Vol. 77, No. 2, 04.2012, p. 244-262.

Research output: Contribution to journalArticle

Petersen, J, Bandeen Roche, KJ, Budtz-Jørgensen, E & Larsen, KG 2012, 'Predicting Latent Class Scores for Subsequent Analysis', Psychometrika, vol. 77, no. 2, pp. 244-262. https://doi.org/10.1007/s11336-012-9248-6
Petersen, Janne ; Bandeen Roche, Karen J ; Budtz-Jørgensen, Esben ; Larsen, Klaus Groes. / Predicting Latent Class Scores for Subsequent Analysis. In: Psychometrika. 2012 ; Vol. 77, No. 2. pp. 244-262.
@article{7a430c0b024743edbcb23ca3740f64c1,
title = "Predicting Latent Class Scores for Subsequent Analysis",
abstract = "Latent class regression models relate covariates and latent constructs such as psychiatric disorders. Though full maximum likelihood estimation is available, estimation is often in three steps: (i) a latent class model is fitted without covariates; (ii) latent class scores are predicted; and (iii) the scores are regressed on covariates. We propose a new method for predicting class scores that, in contrast to posterior probability-based methods, yields consistent estimators of the parameters in the third step. Additionally, in simulation studies the new methodology exhibited only a minor loss of efficiency. Finally, the new and the posterior probability-based methods are compared in an analysis of mobility/exercise.",
keywords = "classification, latent class regression, latent class scores, least squares class, three-step procedure",
author = "Janne Petersen and {Bandeen Roche}, {Karen J} and Esben Budtz-J{\o}rgensen and Larsen, {Klaus Groes}",
year = "2012",
month = "4",
doi = "10.1007/s11336-012-9248-6",
language = "English (US)",
volume = "77",
pages = "244--262",
journal = "Psychometrika",
issn = "0033-3123",
publisher = "Springer New York",
number = "2",

}

TY - JOUR

T1 - Predicting Latent Class Scores for Subsequent Analysis

AU - Petersen, Janne

AU - Bandeen Roche, Karen J

AU - Budtz-Jørgensen, Esben

AU - Larsen, Klaus Groes

PY - 2012/4

Y1 - 2012/4

N2 - Latent class regression models relate covariates and latent constructs such as psychiatric disorders. Though full maximum likelihood estimation is available, estimation is often in three steps: (i) a latent class model is fitted without covariates; (ii) latent class scores are predicted; and (iii) the scores are regressed on covariates. We propose a new method for predicting class scores that, in contrast to posterior probability-based methods, yields consistent estimators of the parameters in the third step. Additionally, in simulation studies the new methodology exhibited only a minor loss of efficiency. Finally, the new and the posterior probability-based methods are compared in an analysis of mobility/exercise.

AB - Latent class regression models relate covariates and latent constructs such as psychiatric disorders. Though full maximum likelihood estimation is available, estimation is often in three steps: (i) a latent class model is fitted without covariates; (ii) latent class scores are predicted; and (iii) the scores are regressed on covariates. We propose a new method for predicting class scores that, in contrast to posterior probability-based methods, yields consistent estimators of the parameters in the third step. Additionally, in simulation studies the new methodology exhibited only a minor loss of efficiency. Finally, the new and the posterior probability-based methods are compared in an analysis of mobility/exercise.

KW - classification

KW - latent class regression

KW - latent class scores

KW - least squares class

KW - three-step procedure

UR - http://www.scopus.com/inward/record.url?scp=84858444368&partnerID=8YFLogxK

UR - http://www.scopus.com/inward/citedby.url?scp=84858444368&partnerID=8YFLogxK

U2 - 10.1007/s11336-012-9248-6

DO - 10.1007/s11336-012-9248-6

M3 - Article

C2 - 23653489

AN - SCOPUS:84858444368

VL - 77

SP - 244

EP - 262

JO - Psychometrika

JF - Psychometrika

SN - 0033-3123

IS - 2

ER -