Psychometric Analysis of the Diagnostic Evaluation of Language Variation Assessment

Abstract

This study investigated the psychometrics of the Diagnostic Evaluation of Language Variation–Screening Test (DELV-S) test using confirmatory factor analysis, item response theory, and differential item functioning (DIF). Responses from 1,764 students in kindergarten through second grade were used in the study, with results indicating that the DELV-S is multidimensional and measures syntactic skills and nonword repetition ability. Item response theory suggested that most items were easy and that the measured skills were most reliable for students who had low language abilities. Standardized effect sizes for DIF suggested small differences existed on syntactic skills between white and minority students. Scores were vertically scaled to produce reference tables to assess performance at specific points in time, as well as growth over time.

Psychometric Analysis of the Diagnostic Evaluation of Language Variation–Screening Test: Extension to Low-Income African American Pre-Kindergarteners

“…With regard to the second goal, significant differences were found between the present study sample and that in Petscher et al (2012). Although this sample of pre-kindergarten children demonstrated longitudinal measurement invariance from the fall to the spring of their school year, they demonstrated only partial measurement invariance with the model of the DELV-S Diagnostic Risk section used in Petscher et al (2012). Because there was not full measurement invariance between that K-2 sample and the present pre-kindergarten sample, separate norms were required for this younger group of test users.…”

“…Results from invariance testing suggest that this group of pre-kindergarten-aged children may represent a different population of test users than those examined by Petscher et al (2012). Perhaps this finding should be expected, as a simple review of the participant characteristics makes it clear that the samples between the studies differed racially and developmentally.…”

“…Summary. Overall, the CFA results indicated that (a) a twofactor structure of syntactic and nonword repetition skills improved fit over a one-factor solution; (b) the two-factor measurement model fit well in a simultaneous estimation from fall to spring; (c) the model parameters for the twofactor model in pre-kindergarten did not strictly conform to those observed by Petscher et al (2012); and (d) by relaxing model thresholds in the two-factor model, it was possible to obtain a solution that demonstrated acceptable fit, retained partial measurement invariance compared with Petscher et al (2012), and was longitudinally invariant. The confluence of these results suggested that item response modeling could be reasonably carried out with separate IRT analyses conducted for each factor so that the unidimensionality assumption could be met.…”

“…This study extended Petscher et al (2012) to include children (primarily African American children who use AAE forms frequently in speech) from low-income households in pre-kindergarten (ages 4-5 years). Notably, African American children who frequently used AAE features in speech were also represented within the larger sample in the aforementioned study; however, 4-year-olds were not, making this study a necessary step in increasing the utility of the DELV-S for describing important characteristics of this student population's language use.…”

“…Responding to this need, Petscher and colleagues (2012) examined the psychometric properties of Part II of the DELV-S, the Diagnostic Risk section. Their diverse sample included children in kindergarten through second grade (ages 5–7 years), where 51% were male, 42% participated in free and reduced price lunch programs, 46% were White, and 38% were African American.…”