Bifactor Structure for the Categorical Chinese Rosenberg Self-Esteem Scale

Area of Science:

• Psychological Measurement
• Psychometrics
• Educational Psychology

Background:

• The Rosenberg Self-Esteem Scale (RSES) is widely used to measure self-esteem.
• Recent research suggests a bifactor model for the RSES latent structure.

Purpose of the Study:

• Investigate differences between bifactor, bifactor negative, and other models for the RSES.
• Examine the impact of treating data as categorical versus continuous.
• Assess the necessity of removing a problematic item in the Chinese RSES.
• Determine the effect on final scoring.

Main Methods:

• Utilized BIC differences and standard model fit indices (RMSEA, CFI, TLI).
• Analyzed data from 1,734 grade 4-6 pupils in Hong Kong.
• Compared model fit and factor loading patterns for categorical and continuous data treatments.

Main Results:

• Strong evidence supports the bifactor model for the 9-item RSES (categorical data).
• Categorical data analysis yielded better factor loading patterns than continuous data.
• Keeping a problematic item had minimal impact on fit indices but caused unexpected negative loadings.
• Method effects in bifactor models were consistently positive, differing from prior research.

Conclusions:

• The bifactor model is well-supported for the RSES, particularly with categorical data.
• The simple sum score is generally appropriate for practical scoring, despite item considerations.
• Understanding factor loadings and method effects is crucial for accurate RSES interpretation.