An Information Manifold Perspective for Analyzing Test Data

Area of Science:

• Psychometrics
• Information Theory
• Educational Measurement

Background:

• Current psychometric models for test data analysis have limitations.
• A need exists for additive scale measures of information.
• Existing models may not fully utilize all available information, including distractors.

Purpose of the Study:

• To propose modifications to psychometric models using information theory.
• To introduce an additive scale measure of information based on manifold geometry.
• To evaluate item performance, inter-item dependency, and test taker knowledge more effectively.

Main Methods:

• Development of a one-dimensional space curve or curved surface manifold for information measurement.
• Utilizing arc length along the manifold as an additive metric with a defined zero and bit unit.
• Incorporating all item information, including distractors, into the analysis.
• Comparison with the item response theory nominal model using large-scale college admissions test data.

Main Results:

• The proposed model generates an additive scale measure of information invariant across indexing systems.
• The arc length of the manifold quantifies the 'scope' of a test or item.
• Test taker performance is represented by a position along the curve.
• The information theory perspective offers novel methods for assessing items and knowledge.

Conclusions:

• The test information manifold perspective provides a powerful new framework for psychometric analysis.
• This approach allows for a more comprehensive evaluation of tests and individual performance.
• Information theory offers innovative ways to assess item characteristics and learner knowledge.