Jove
Visualize
Contact Us
JoVE
x logofacebook logolinkedin logoyoutube logo
ABOUT JoVE
OverviewLeadershipBlogJoVE Help Center
AUTHORS
Publishing ProcessEditorial BoardScope & PoliciesPeer ReviewFAQSubmit
LIBRARIANS
TestimonialsSubscriptionsAccessResourcesLibrary Advisory BoardFAQ
RESEARCH
JoVE JournalMethods CollectionsJoVE Encyclopedia of ExperimentsArchive
EDUCATION
JoVE CoreJoVE BusinessJoVE Science EducationJoVE Lab ManualFaculty Resource CenterFaculty Site
Terms & Conditions of Use
Privacy Policy
Policies

Related Experiment Videos

Inferring the structure of latent class models using a genetic algorithm.

Han L J van der Maas1, Maartje E J Raijmakers, Ingmar Visser

  • 1Department of Psychology, University of Amsterdam, Roetersstraat 15, 1018 WB Amsterdam, The Netherlands. hvandermaas@fmg.uva.nl

Behavior Research Methods
|September 21, 2005
PubMed
Summary
This summary is machine-generated.

Related Concept Videos

You might also read

Related Articles

Articles linked to this work by shared authors, journal, and citation graph.

Sort by
Same author

A reciprocal model of practice and skill: Navigating between dropout and expertise.

Psychological review·2026
Same author

Polarization in increasingly connected societies.

Physical review. E·2026
Same author

Are Semantic Representations Stable? A Bayesian Framework Applied to the Study of Quantifier Meaning.

Computational brain & behavior·2026
Same author

What 5000 babies can tell us about developing minds and how to study them.

Communications psychology·2026
Same author

From night-to-night and from person-to-person: dynamic phenomena of insomnia.

Sleep advances : a journal of the Sleep Research Society·2026
Same author

A theory-construction methodology for network theories in psychology.

Psychological methods·2026
Same journal

The Rhythm Reproduction Task for children: A psychometric examination.

Behavior research methods·2026
Same journal

QualGames: A Qualtrics implementation and a database of behavioral game theory tasks.

Behavior research methods·2026
Same journal

PyLossless: A non-destructive EEG processing pipeline.

Behavior research methods·2026
Same journal

"It makes sense to me": Examining data file column names in the visual cognitive literature.

Behavior research methods·2026
Same journal

Planned missingness in intensive longitudinal studies: Extensions and comparisons of multiform designs.

Behavior research methods·2026
Same journal

A validity-guided workflow for robust large language model research in psychology.

Behavior research methods·2026
See all related articles

This study introduces a novel genetic algorithm for latent class analysis, optimizing model structure and parameters simultaneously. This approach enhances model fitting compared to traditional methods, improving discoverability of optimal latent class structures.

Area of Science:

  • Statistics
  • Computational Statistics
  • Psychometrics

Background:

  • Traditional latent class analysis (LCA) relies on Expectation-Maximization (EM) or Newton-Raphson algorithms for parameter optimization.
  • Model structure in LCA, including the number of classes and constraints, is typically predetermined by theoretical considerations.
  • Model comparison often involves goodness-of-fit indices, but identifying the optimal structure can necessitate extensive, time-consuming searches.

Purpose of the Study:

  • To develop and present a novel method for simultaneously inferring latent class analysis (LCA) model structure and optimizing parameter values.
  • To introduce a genetic algorithm (GA) approach for LCA that utilizes any goodness-of-fit index as a fitness criterion.
  • To demonstrate the efficacy of the GA-based LCA method in finding optimal model structures.

Related Experiment Videos

Main Methods:

  • A genetic algorithm (GA) was developed to perform simultaneous model structure inference and parameter optimization in latent class analysis.
  • The GA employs a chosen goodness-of-fit index as its fitness function to guide the search for optimal models.
  • The proposed method was evaluated using existing datasets from the literature to compare its performance against established LCA models.

Main Results:

  • The genetic algorithm approach successfully identified latent class models that fit the data as well as or better than models derived from traditional methods.
  • The GA method offers a more efficient way to explore and determine optimal model structures compared to lengthy manual searches.
  • Solutions obtained through the GA method showed robustness, potentially mitigating issues related to the choice of initial parameter values.

Conclusions:

  • The presented genetic algorithm provides an effective and efficient alternative for latent class analysis, enabling simultaneous optimization of model structure and parameters.
  • This novel approach facilitates the discovery of superior latent class models by integrating structure search with parameter estimation.
  • The GA-based method holds significant potential for advancing the application and interpretability of latent class models in various research fields.