The Effect of Including or Excluding Students with Testing Accommodations on IRT Calibrations.

This study aimed to determine the degree to which the inclusion of accommodated students with disabilities in the calibration sample affects the characteristics of item parameters and the test results. Investigated were effects on test reliability, item fit to the applicable item response theory (IRT) model, item parameter estimates, and students' scores. Data were obtained from a statewide standards-based assessment program for reading and writing in grades 4 and 7 and in mathematics and science for grade 8. One data set comprised all tested students, including those who tested with accommodations. The other set was only those students who tested without accommodations. Differential item functioning was studied for accommodated groups. The percent of students who received accommodations varied by grade, ranging between 4.2% and 8.5% of the tested population, with higher percentages in the lower grades. The accommodation used most often was extended time followed by oral presentation. There were notable effects of including or excluding students with accommodations, and statistically significant differences between the inclusive and exclusive parameter estimates indicate that choice of calibration sample does not have a significant effect on the calibration results. Effects of these differences on the mean fit of items to the IRT model are observable but very small. Effects of differences in the item parameters on test results are less notable. Overall, results show few negative effects of calibration inclusiveness. (Contains 8 tables and 16 references.) (SLD)