A Reliability Generalization Study of Coefficient Alpha for the UCLA Loneliness Scale

The goal of this investigation was the development of an Inconsistency scale (ICN–SF) for the Personality Assessment Inventory–Short Form (PAI–SF). In Study 1, 503 inpatient profiles were randomly assigned to a derivation or cross-validation sample. Ten correlated item pairs were identified using the derivation sample and placed on the ICN–SF. Psychometric properties of the ICN–SF total scores were comparable in the derivation and cross-validation samples. Total ICN–SF scores in both samples were significantly lower than scores obtained from computer-generated random samples. Diagnostic efficiency statistics are reported using multiple cut-off scores at various base rate estimates. ICN–SF scores greater than 8 reasonably balanced sensitivity and specificity rates. This cutoff correctly classified 92% of the random protocols and inaccurately classified 9% of the patient protocols in Study 1. In Study 2, PAI–SF scores from 627 forensic and civil inpatients produced similar results, effectively identifying cases with elevated scores on the full-form Inconsistency scale. Overall the results of both studies suggest that the ICN–SF can aid examiners in assessing for inconsistent responding.