How to Compare Psychometric Factor and Network Models.

Area of Science:

• Psychology
• Psychometrics
• Statistical Modeling

Background:

• Dr. Dennis John McFarland's research focused on comparing psychometric networks and latent variable models.
• The study serves as a commentary on McFarland's final work, continuing his statistical investigations.
• Previous work often utilized structural equation modeling (SEM) in standard software.

Purpose of the Study:

• To statistically compare psychometric network models with latent variable models.
• To evaluate the suitability of standard SEM software for this comparison.
• To reanalyze intelligence data and demonstrate a novel comparison method.

Main Methods:

• Utilized structural equation modeling (SEM) for initial analysis.
• Evaluated the penta-factor model of intelligence.
• Employed the R package Psychonetrics for a novel network and latent variable model comparison.
• Reanalyzed Wechsler Adult Intelligence Scale data.

Main Results:

• Standard SEM software is unsuitable for comparing psychometric networks and latent variable models.
• The penta-factor model of intelligence was found to be nonidentified and of limited value.
• The R package Psychonetrics effectively facilitates the comparison of network and latent variable models.
• Intelligence network models demonstrated replication across standardization samples.

Conclusions:

• Standard SEM software limitations necessitate alternative approaches for comparing psychometric networks and latent variable models.
• A network interpretation of general intelligence is supported by the findings.
• The replicability of intelligence networks across samples is a significant empirical finding.