Modeling of Item Response Functions Under the D-Scoring Method.

Area of Science:

• Educational Measurement
• Psychometric Methods
• Large-Scale Assessments

Background:

• The D-scoring method (DSM) is increasingly used in educational and psychological measurement.
• Previous DSM applications utilized logistic regression models (LRMs) for item response functions (IRFs).
• LRMs exhibit underestimation issues for item true scores at the upper D-scale range, particularly for difficult items.

Purpose of the Study:

• To introduce novel rational function models (RFMs) for estimating item response functions within the D-scoring method framework.
• To address the underestimation and psychometric inaccuracies associated with logistic regression models in DSM.
• To offer simpler and more accurate alternatives to complex inverse-regression adjustments.

Main Methods:

• Development and proposal of one-parameter (RFM1), two-parameter (RFM2), and three-parameter (RFM3) rational function models.
• Evaluation of the proposed RFMs using simulated data.
• Validation of the RFMs with real-world assessment data.

Main Results:

• The proposed rational function models (RFMs) provide more accurate estimations of item true scores compared to logistic regression models (LRMs).
• RFMs effectively resolve the underestimation issues at the top of the D-scale observed with LRMs.
• The new models offer improved accuracy in standard error estimates and overall psychometric properties for DSM.

Conclusions:

• Rational function models (RFMs) offer a superior alternative to logistic regression models for item response function estimation in the D-scoring method (DSM).
• The proposed RFMs enhance the precision and reliability of D-scores in educational and psychological assessments.
• These models simplify complex psychometric adjustments while maintaining high accuracy for large-scale assessments.