How to reduce the number of rating scale items without predictability loss?

Area of Science:

• Psychometrics
• Quantitative Psychology
• Research Methodology

Background:

• Rating scales are crucial for collecting data on qualitative constructs.
• Existing scales can be lengthy, leading to data redundancy and collection inefficiencies.

Purpose of the Study:

• To present an innovative method for reducing the number of items in rating scales.
• To demonstrate that item reduction does not compromise data predictability or scale reliability.

Main Methods:

• The study employed the area under the receiver operator curve (AUC ROC) method for item reduction.
• Statistical validation was performed using the Graded Response Model (GRM) and Confirmatory Factor Analysis (CFA).

Main Results:

• The AUC ROC method reduced rating scale items by over 70% (from 21 to 6 variables).
• The Graded Response Model confirmed the reduced scale's ability to differentiate between high and middle scores.
• Confirmatory Factor Analysis verified that scale reliability was maintained post-reduction.

Conclusions:

• The AUC ROC method offers an effective approach to data reduction in rating scales.
• This method enhances research efficiency by minimizing unnecessary data collection without sacrificing analytical power or reliability.