Related Experiment Videos
The use of causal indicators in covariance structure models: some practical issues
1Department of Psychology, Ohio State University, Columbus 43210.
Psychological Bulletin
|November 1, 1993
Summary
This study explores composite variables and causal indicators in covariance structure models, highlighting potential identification issues and unintended zero correlations. It offers remedies for evaluating models with causal indicators.
Area of Science:
- Psychometrics
- Statistical Modeling
Background:
- Conventional covariance structure models define indicators as linear functions of latent variables plus error.
- An alternative approach defines constructs as linear functions of indicators (causal indicators) plus error, creating composite variables.
Purpose of the Study:
- To investigate the implications of using composite variables and causal indicators in statistical models.
- To identify and address potential problems with parameter identification and implied correlations.
Main Methods:
- The study demonstrates phenomena using a specific example.
- General principles underlying the use of causal indicators are discussed.
Main Results:
- Composite variables can lead to parameter identification problems in covariance structure models.
- Causal indicators may imply zero correlations among measured variables, requiring additional parameters.
Conclusions:
- Models with causal indicators can present unique challenges in statistical analysis.
- Remedies are provided to facilitate the evaluation of models incorporating causal indicators.