ASSESSMENT OF INFORMATION LOSS IN SCORING MONOTONE ITEMS

Area of Science:

• Psychometrics
• Statistical analysis
• Measurement theory

Background:

• Dichotomous items are common in psychometric assessments.
• Standard scoring procedures for such items assume unidimensionality.
• Deviations from a perfect Guttman Scale present interpretation challenges.

Purpose of the Study:

• To differentiate between information loss and information confusion in scoring dichotomous items.
• To evaluate two interpretations of scoring procedures when Guttman Scale assumptions are violated.
• To propose appropriate methods for analyzing nonunidimensional item sets.

Main Methods:

• Development of measures for information loss under two distinct interpretations.
• Testing these measures using Monte Carlo simulations.
• Analysis of scoring procedures for monotone, dichotomous items.

Main Results:

• Interpretation A (information loss) is less appropriate when Guttman Scale assumptions fail.
• Interpretation B (information confusion) accurately reflects the impact of nonunidimensionality on scoring.
• Monte Carlo results support the distinction between the two interpretations.

Conclusions:

• Nonunidimensionality in item sets leads to information confusion, not loss, under usual scoring.
• Interpretation B provides a more accurate framework for understanding scoring issues with Guttman Scales.
• The developed measures aid in assessing the impact of scale properties on data interpretation.