LSAC Resources - Research

Bayesian Estimation Methods for Multidimensional Models for Discrete and Continuous Responses with a Structure on the Item and Person Parameters (RR 07-03)

by Cees A. W. Glas and Oksana Korobko, University of Twente, Enschede, The Netherlands

Executive Summary

In previous Law School Admission Council reports, models for the simultaneous analysis of discrete (correct versus incorrect) and continuous item responses were developed. The current extension of that research focuses on the prediction of item and person parameters from covariates.

The method is illustrated with examples of the analysis of the grades for national school-leaving examinations at the end of secondary education in the Netherlands. In this example, as in one of the previous reports, the interaction between the students' patterns and levels of proficiency and the choice of examination subjects is modeled by extending the basic model for the responses with a model for the choice of the examination subjects that were elective for the students. Using this setup, two research questions are investigated: (a) How much of the variance in the ability parameters for the person is attributable to the schools? and (b) How much of this variance is attributable to gender?

With respect to research question (a), the schools contributed to the students' ability in Language and Economy, but not to their ability in Science. With respect to research question (b), the proportions of variance explained by gender were highest for the ability dimension associated with Language and Science. Further, the average score for ability associated with the Language dimension was higher for female students, while their average on the choice dimension was lower.

• Full Research Report (PDF)