Lin, P.-Y., & Lin, Y.-C. (2016). Examining accommodation effects for equity by overcoming a methodological challenge of sparse data . Research in Developmental Disabilities , 51–52 , 10–22.

Journal Article

Lin, P.-Y., & Lin, Y.-C. (2016). Examining accommodation effects for equity by overcoming a methodological challenge of sparse data. Research in Developmental Disabilities, 51–52, 10–22.


[no issue, just volumes 51-52]


Breaks during testing; Dictated response; Dictated response (scribe); Dictated response (speech recognition system); Electronic administration; Electronic administration; Electronic administration; Emotional/Behavioral disability; Extended time; High school; International (non-U.S.); Language arts; Learning disabilities; Multiple accommodations; Multiple disabilities; Oral delivery; Oral delivery of directions only; Recorded delivery (audio or video); Specialized setting; Technological aid; Word processing (for writing)




For this application of data adjustment methods during log-linear and odds ratio analyses, the researchers investigated the impact of 31 sets of accommodations, including individual accommodations and combinations of accommodations—also referred to as bundled accommodations. The accommodations included assistive devices and technology for communication, breaks (supervised and periodic), computer or word processor for responses, extended time, preferential seating within the regular classroom, prompts for students with severe attention problems, scribe, setting, and verbatim reading of writing prompts and tasks. As a reference point, students with learning disabilities (only) completing the provincial assessment (in Canada) were documented as having used 574 different sets of accommodations, either individual supports or combinations of them.


Grade 10 students who took the Ontario [province of Canada] Secondary School Literacy Test (OSSLT) were sampled to include 945 students with emotional or behavioral disabilities, 12,690 students with learning disabilities, and 922 students with multiple disabilities. The total number of participants consisted of the 14,499 students with these disabilities who used at least one accommodation during the assessment.

Dependent Variable

The extant data set from 2012-2013 for the grade 10 Ontario Secondary School Literacy Test (OSSLT), a provincial graduation assessment in Canada, was sampled for this secondary analysis and non-experimental study. The OSSLT, an English literacy assessment, consisted of multiple-choice and constructed-response items. Specifically, whether or not students achieved at or above proficiency levels—that is, levels 3 and 4—was documented. The researchers noted that the approach was not strictly experimental, as the samples of students with these disabilities were not matched based on backgrounds, learning characteristics, or difficulties.


The odds ratio results demonstrated that the three groups of students with disabilities receiving the combinations of: (a) computer administration with extended time, (b) computer administration with specialized setting, and (c) computer administration with both extended time and specialized setting, all "had a higher chance to meet the [literacy] standards" (p. 20) than their peers receiving no accommodations, and also than their peers receiving other accommodations. The students with learning disabilities had the most pronounced differences. The researchers cautioned that the matter of accommodations potentially affecting construct validity might be of concern, noting that the OSSLT had both multiple-choice and constructed-response items; the potential that students' responses composed using assistive technology—such as speech-to-text software—could affect the test construct could not be evaluated. The researchers also reported that one of the odds ratio data adjustment methods was found not to be useful, while another approach demonstrated usefulness, in addressing the problem of data sparsity, particularly for some less commonly used accommodations. Specifically, the treatment arm correction method from a previous study (Sweeting, Sutton, & Lambert, 2004) was found not useful in the current study; on the other hand, the application in the current study of the log-linear analysis and adjusted odds ratio method from a different study (Diamond, Bax, & Kaul, 2007) was useful.