Related Experiment Videos
Being inconsistent about consistency: when coefficient alpha does and doesn't matter
1Baycrest Centre for Geriatric Care, Department of Psychiatry, University of Toronto, Ontario, Canada. dstreiner@klaru-baycrest.on.ca
Journal of Personality Assessment
|May 24, 2003
Summary
Classical test theory applies to scales measuring underlying constructs, not indexes where items define the construct. This distinction impacts item selection and reliability analysis for psychological measures.
Area of Science:
- Psychometrics
- Psychological Measurement
- Classical Test Theory
Background:
- Classical test theory emphasizes internal consistency (Cronbach's alpha, interitem correlations) for scale reliability.
- This principle, however, may not universally apply to all types of measurement instruments.
Purpose of the Study:
- To differentiate between "scales" (effect indicator models) and "indexes" (causal indicator models) in psychometric assessment.
- To explore the implications of this distinction for item selection, item relationships, and reliability estimation.
Main Methods:
- Distinction based on Bollen and Lennox (1991) framework differentiating "effect indicators" from "causal indicators".
- Conceptual analysis of measurement models and their alignment with classical test theory assumptions.
Main Results:
- "Scales" with effect indicators align with classical test theory's internal consistency requirements.
- "Indexes" with causal indicators do not meet these traditional criteria.
Conclusions:
- Measurement instruments should be classified as either scales or indexes based on their item-indicator relationship.
- Appropriate statistical methods for reliability assessment differ significantly between scales and indexes.