Hierarchical Cluster Analysis And The Internal Structure Of Tests

Area of Science:

• Psychometrics
• Data Analysis
• Psychological Measurement

Background:

• Scale construction is crucial for psychological measurement.
• Traditional methods like factor analysis have limitations with large item pools.
• Assessing the psychometric adequacy of scales is essential.

Purpose of the Study:

• To demonstrate hierarchical cluster analysis as an effective method for scale construction.
• To introduce coefficient beta as a novel measure for internal consistency reliability.
• To compare hierarchical clustering with factor analytic techniques for scale development.

Main Methods:

• Hierarchical cluster analysis was applied to item pools.
• Psychometric adequacy of potential scales was tested.
• Coefficient beta, a new measure of internal consistency reliability (worst split-half reliability), was used to assess scale adequacy.
• Comparisons were made with conventional factor analytic techniques.

Main Results:

• Hierarchical cluster analysis proved effective in forming scales.
• Higher-order scales were formed when they demonstrated greater adequacy than sub-scales.
• Coefficient beta provided a new criterion for assessing scale adequacy.
• Hierarchical clustering, guided by psychometric decisions, showed advantages over factor analysis for large item pools.

Conclusions:

• Hierarchical cluster analysis offers a robust approach to scale construction.
• Coefficient beta is a valuable new metric for evaluating internal consistency reliability.
• This psychometrically-informed clustering method is superior to factor analysis for developing scales from extensive item sets.