Roxbury, T. L. (2010). A psychometric evaluation of a state testing program: Accommodated versus non-accommodated students (Publication No. 3403710) [Doctoral dissertation, The University of North Carolina at Greensboro]. ProQuest Dissertations and Theses Global. https://www.proquest.com/docview/365998592

The University of North Carolina (Greensboro, NC); ProQuest document ID: 365998592

The accommodations examined in this study were those allowed naturalistically by the state; for the purposes of this study, the accommodations were grouped together in the analyses, and were not examined separately. The frequency of accommodations use was reported for each content area. The most commonly used accommodations included small group administration, extended time, and checking comprehension of directions, among others.

The participants were 15,274 grade 8 students in an unspecified state in the Northeast (U.S.) who completed statewide testing. Of these, 2,508 students had disabilities of all categories naturally occurring in the population of the state. Data were reported on demographic characteristics, including sex, ethnicity, and SES, among others. For the mathematics test, 2,227 students were tested with accommodations; for the reading test, 2,221 students used accommodations; for the science test, 2,197 students used accommodations.

The dependent variable was the statewide testing for mathematics, reading, and science at the grade 8 level. While the tests included multiple-choice and constructed-response items, this study examined only the multiple-choice items and responses.

The results of the analyses offer conclusions on the comparison of performance between accommodated and non-accommodated groups, and offer information about the functioning of the test regarding fairness for students with disabilities. First of all, students who used accommodations scored significantly lower than students who did not use accommodations. Secondly, examination of the tests (although not the items, which were not released and not available) yields that each content area test was unidimensional—which indicates that accommodated and non-accommodated students were not being somehow tested on different constructs—that is, accommodations did not change the constructs. Finally, DIF analysis indicated that some items exhibited differential item functioning, although lacking the specific item texts meant that possible reasons could not be offered for the basis of the bias. Limitations of the study were reported, and future research possibilities were suggested.