Digit Span Subscale Scores May Be Insufficiently Reliable for Clinical Interpretation: Distinguishing Between Stratified Coefficient Alpha and Omega Hierarchical. (December 2019)

• Record Type: Journal Article
• Authors: Gignac, Gilles E.; Reynolds, Matthew R.; Kovacs, Kristof
• Is Part Of: Assessment. Volume 26:Number 8(2019)
• Page Start: 1554
• Page End: 1563
• Publication Date: 2019-12
• Subjects: memory -- digit span backward -- digit span forward -- digit span sequencing -- reliability; Psychological tests -- Periodicals; Psychodiagnostics -- Periodicals; Personality tests -- Periodicals; 150.287
• DOI: 10.1177/1073191117748396
• Languages: English
• ISSNs: 1073-1911

Abstract

The Digit Span subscale (Digit Span Forward, Backward, and Sequencing combined composite) internal inconsistency reliability has been reported at .93, based on a coefficient known as stratified coefficient alpha. With accessible examples, we demonstrate that stratified coefficient alpha can deviate substantially from a model-based internal consistency reliability that represents an underlying dimension, that is, omega hierarchical. Next, we simulated item-level Digit Span subscale data to correspond very closely to the Wechsler Adult Intelligence Scale–fourth edition normative sample. Based on omega hierarchical, we estimated the internal consistency reliability associated with the Digit Span subscale scores at .74. In light of the results, clinicians are cautioned against interpreting Digit Span subscale scores. Instead, interpretations should probably be restricted to the Digit Span Forward, Digit Span Backward, and Digit Span Sequencing test scores. Finally, we offer suggestions for improvement to achieve higher levels of Digit Span subscale score reliability.