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

Scaled test statistics and robust standard errors for non-normal data in covariance structure analysis: a Monte Carlo

C P Chou1, P M Bentler, A Satorra

  • 1Department of Preventive Medicine, University of Southern California, Alhambra 91803-1358.

The British Journal of Mathematical and Statistical Psychology
|November 1, 1991
PubMed
Summary

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

Invariant Standardized Estimated Parameter Change for Model Modification in Covariance Structure Analysis.

Multivariate behavioral research·2016
Same author

A NEW MATRIX FOR THE ASSESSMENT OF FACTOR CONTRIBUTIONS.

Multivariate behavioral research·2016
Same author

Brief Report: An Additional Minimal Transformation To Orthonormality.

Multivariate behavioral research·2016
Same author

The Relationship Of Personality Structure To Patterns Of Adolescent Substance Use.

Multivariate behavioral research·2016
Same author

Interrelations Among Models For The Analysis Of Moment Structures.

Multivariate behavioral research·2016
Same author

Longitudinal Analysis Of The Role Of Peer Support, Adult Models, And Peer Subcultures In Beginning Adolescent Substance Use: An Application Of Setwise Canonical Correlation Methods.

Multivariate behavioral research·2016

Maximum Likelihood (ML) and Asymptotic Distribution Free (ADF) statistics can be biased under non-normal data. A scaled test statistic and robust standard errors offer improved performance in covariance structure analysis.

Area of Science:

  • Statistics
  • Psychometrics
  • Quantitative Psychology

Background:

  • Maximum Likelihood (ML) statistics in covariance structure analysis are known to be biased under severe non-normality.
  • Asymptotic Distribution Free (ADF) estimation and scaled test statistics are proposed alternatives to mitigate these biases.

Purpose of the Study:

  • To compare the performance of scaled test statistics and robust standard errors against ML and ADF methods.
  • To evaluate the robustness of these statistical approaches under various non-normal conditions for two distinct models.

Main Methods:

  • Comparison of scaled test statistics, robust standard errors, ML, and ADF methods.
  • Evaluation across two models under several non-normal data conditions.

Main Results:

Related Experiment Videos

  • ML and ADF test statistics showed variable performance across models; ADF performed worst overall.
  • The scaled test statistic generally outperformed the ML test statistic.
  • Robust and ADF standard errors provided better estimates of sampling variability than ML standard errors, which were often downward biased.
  • ML statistics demonstrated robustness under symmetric, platykurtic, or non-symmetric, zero-kurtotic distributions.

Conclusions:

  • Scaled test statistics and robust standard errors present a viable alternative to ML and ADF methods, particularly under non-normal conditions.
  • Careful consideration of data distribution is crucial when selecting statistical methods for covariance structure analysis.