LSAT Dimensionality Analysis for the December 1991, June 1992, and October 1992 Administrations (SR-95-05)
by Jeff Douglass, Hae-Rim Kim, University of Illinois at Urbana-Campaign; Louis Roussos, Law School Admission Council; William Stout, and Jinming Zhang, University of Illinois at Urbana-Champaign

Executive Summary

For the purposes of this report, the dimensions of a test may be thought of as the number of statistically detectable constructs that the test is measuring. A dimensionality analysis of a test involves determining the number of dimensions being measured by the test, the nature of these dimensions, and the degree to which the dimensions are correlated. Thus, an assessment of the dimensionality structure of the Law School Admissions Test (LSAT) can a play an important role in maintaining and possibly improving the high psychometric quality of the LSAT. To this end an extensive dimensionality analysis has been conducted on the December 1991, June 1992, and October 1992 administrations of the LSAT using nonparametric tools recently developed by the authors of this report. Recent advances in the development of nonparametric tools offered the potential of powerful detection of multidimensionality while avoiding the stronger (possibly incorrect) assumptions of parametric tools.

The results of the analysis indicate that the LSAT displays a moderate amount of multidimensionality, appearing to have two dominant moderately correlated dimensions, one corresponding to the Analytical Reasoning (AR) items and one corresponding to the combined Logical Reasoning (LR) and Reading Comprehension (RC) items. These results are in agreement with the results of previously conducted parametric dimensionality analyses. The current analyses also revealed several secondary dimensions that also had been implicated by the results of previous parametric analyses. For example, our results indicate that the LR and RC sections are dimensionally distinct (though highly correlated) and that the passage-based AR and RC item sets are dimensionally distinct from each other. Finally, unlike the previous parametric analyses, our analyses also indicate that the LR items, while displaying by far the lowest level of multidimensionality of any of the item types, appear to have several weak and less easily detectable secondary dimensions. The most reliable of these weaker dimensions is an end-of-section possible speededness effect. Further analysis of these weaker dimensions is an area of future investigation because these dimensions may play a role in cognitive diagnosis or in Differential Item Functioning (DIF).

LSAT Dimensionality Analysis for the December 1991, June 1992, and October 1992 Administrations (SR-95-05)

Research Report Index