Date of Award

2024

Publication Type

Thesis

Degree Name

M.A.

Department

Psychology

Supervisor

Laszlo Erdodi

Abstract

Background: Performance validity tests (PVTs) assess whether an examinee’s performance reflects their true ability. This thesis was designed to evaluate whether the strategic combination of independent EVIs within the Wechsler Adult Intelligence Scale-IV (WAIS-IV) improved classification accuracy over individual WAIS-IV EVIs. Given the ubiquity of the WAIS-IV in assessment, it is an excellent candidate to be evaluated as a multivariate PVT. Methods: Archival data was examined from a clinical sample of 84 adults referred for evaluation of various psychiatric and neurological disorders and a second medicolegal sample of 110 adults mostly referred for evaluation of traumatic brain injury. The classification accuracy of individual subtests and other EVIs nested within the WAIS-IV were evaluated against psychometrically defined non-credible performance (i.e., criterion PVTs). Multivariate models were developed based on EVIs with strong classification accuracy against these criterion PVTs. Results: Two multivariate WAIS-IV PVTs were developed using sample-specific cut-offs. A seven-component model containing the Vocabulary, Matrix Reasoning, Block Design, Digit Span (DS), Coding (CD), Symbol Search (SS), and CD/SSRaw EVIs (i.e., WAIS-VI-7) and a four-component model made up of the DS, CD, SS, and CD/SSRaw EVIs (i.e., WAIS-VI-4). Failure on ≥3 on the WAIS-VI-7 yielded good specificity (.89-.92) but low sensitivity (.30-.33) in both groups. Failure on ≥2 on the WAIS-VI-4 in the clinical sample had .91 specificity and .30 sensitivity, while a more conservative ≥3 cut-off was necessary in the medicolegal sample (.98 specificity, .13 sensitivity). Overall, the multivariate WAIS-IV models had improved classification accuracy compared to the individual EVIs. Conclusion: The present study provided evidence that the WAIS-VI-7 and WAIS-VI-4 multivariate models have the potential to become effective tools for detecting noncredible responding. However, their low sensitivity remains a liability.

Share

COinS