Unipolar IRT and the Author Recognition Test (ART)

Area of Science:

• Psychometrics
• Educational Measurement
• Psychological Assessment

Background:

• Item Response Theory (IRT) is commonly used to assess measurement error in tests.
• Traditional IRT models assume normally distributed traits and bipolar constructs, using cumulative normal/logistic link functions.
• Unipolar IRT models are increasingly recognized as more suitable for certain personality and psychopathology constructs.

Purpose of the Study:

• To extend the application of unipolar IRT models to a recognition task measure, specifically the Author Recognition Test (ART).
• To investigate the implications of choosing between unipolar and bipolar IRT models for quantifying and interpreting measurement error in the ART.
• To compare measurement error patterns derived from unipolar versus bipolar IRT analyses of the ART.

Main Methods:

• Application of unipolar Item Response Theory (IRT) models.
• Analysis of several example datasets from various versions of the Author Recognition Test (ART).
• Comparison of results from unipolar IRT analyses with prior bipolar IRT analyses of the ART.

Main Results:

• The choice between unipolar and bipolar IRT modeling significantly impacts the quantification and interpretation of measurement error in the ART.
• In contrast to bipolar IRT analyses, unipolar IRT revealed that measurement error in the ART is minimized at low levels of latent print exposure.
• Measurement error in the ART increases as latent print exposure increases under the unipolar IRT framework.

Conclusions:

• Unipolar IRT modeling provides a distinct understanding of measurement error for the Author Recognition Test (ART), differing substantially from bipolar IRT.
• The findings suggest that unipolar IRT may be more appropriate for constructs like print exposure, where the response process differs from typically bipolar traits.
• Consideration of unipolar IRT is recommended for other constructs and measures, such as vocabulary or specialized knowledge, that share similar response processes.