skip to main content skip to footer

An Empirical Investigation of One Variable Section Pre-Equating

Wightman, Linda F. (Leary); Wightman, Lawrence E.
Publication Year:
Report Number:
ETS Research Report
Document Type:
Page Count:
Subject/Key Words:
Equated Scores, Mathematical Models, Scoring, Section Pre-Equating (Tests), Test Interpretation


Section Pre-Equating (SPE) is a method used to equate test forms that consist of multiple separately-timed sections. The unique contribution of SPE is that it does not require examinees to take two complete forms of the test. Instead, all of the old form and one or two sections of the new form are administered to each examinee. Missing data techniques are employed to estimate the necessary equating parameters. When a test has two variable sections, the estimation is fairly straightforward. When a test includes only one variable section, there is no simple way to obtain an estimate of the correlation between pairs of sections from the new test because no one group takes a pair of sections. This study was designed to utilize empirical data to evaluate several methods for obtaining reasonable values for these correlations. The methods for obtaining correlations that are evaluated are borrowing correlations from another population who have taken the same test at a different administration, borrowing correlations from another parallel test taken by the same population, and imputing the unknown partial correlations. This study explored the effect on equating of using each of these methods. The criterion for evaluating the one-variable-section test equatings was the equating obtained using a two-variable-section model. Comparisons of the converted scores obtained using the three different methods for estimating correlations under a one-variable-section test model were consistent with one another. Results from this study also demonstrated that equating results obtained from a one-variable-section test model were very consistent with those obtained from a two-variable-section test model. (43pp.)

Read More