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

Fitting the factor analysis model in lI norm.

Nickolay T Trendafilov1

  • 1University of the West of England, Bristol, UK. nickolay.trendafilov@uwe.ac.uk

The British Journal of Mathematical and Statistical Psychology
|June 23, 2005
PubMed
Summary
This summary is machine-generated.

This study introduces a robust method for exploratory factor analysis, using a resistant discrepancy measure instead of least squares. This approach improves the estimation of factor analysis model parameters for better data analysis.

Related Experiment Videos

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

Clustered Common Factor Exploration in Factor Analysis.

Psychometrika·2019
Same author

Some Mathematical Properties of the Matrix Decomposition Solution in Factor Analysis.

Psychometrika·2017
Same author

Sparse Exploratory Factor Analysis.

Psychometrika·2017
Same author

Sparse Versus Simple Structure Loadings.

Psychometrika·2014
Same author

Independent component analysis as a rotation method: a very different solution to Thurstone's box problem.

The British journal of mathematical and statistical psychology·2005

Area of Science:

  • Statistics
  • Psychometrics
  • Data Analysis

Background:

  • Exploratory factor analysis (EFA) traditionally uses least squares for model fitting, which can be sensitive to outliers.
  • Fitting EFA models involves complex matrix optimization with constraints on parameter structures, such as positive definiteness.

Discussion:

  • This research proposes a novel approach by replacing the least squares goodness-of-fit function with a resistant discrepancy measure, specifically a smooth approximation of the l1 norm.
  • The study employs a projected gradient method, well-suited for matrix optimization problems, to address the challenges of fitting the EFA model to sample correlation matrices.
  • Two distinct reparameterizations of the EFA model are explored to enhance the fitting procedure.

Key Insights:

  • The projected gradient approach ensures the preservation of parameter structures during optimization, leading to more reliable factor analysis solutions.
  • The proposed method demonstrates globally convergent procedures for the simultaneous estimation of factor analysis matrix parameters.
  • Numerical examples validate the effectiveness of the developed algorithms and the quality of the resulting factor analysis solutions.

Outlook:

  • This robust methodology offers a valuable alternative for EFA, particularly in datasets with potential outliers or complex structures.
  • Further research could explore the application of this resistant fitting approach to other multivariate statistical models.
  • The development of efficient algorithms for parameter estimation in constrained matrix optimization problems remains a key area for statistical advancement.