Psychometric evaluation of revised Task-Related Worry Scale (TRWS-R): A Mokken model analysis

Task-related worries can be understood as an inherent component of an anxious state and stress response. Under evaluating conditions (e.g. cognitive testing), these worries, due to cognitive interference they create, may have undesirable effects on a cognitive performance at hand. Since cognitive interference has been documented to affect a broad spectrum of cognitive performance (Hembree, 1988), development of a method for its assessment is required. For this purpose we modified a part of the original Cognitive Interference Questionnaire (Sarason et al., 1986) in order to create the revised Task- Related Worry Scale (TRWS-R) and investigated its psychometric properties. Data from two hundreds of participants (72 male, 139 female; age ranging from 18 to 24) were obtained to inspect the modified scale’s properties on Slovak sample. After the scale was reformulated and shortened, the resulting set of eight items was subjected for examination of internal consistency (Cronbach'salpha, Revelle’sbeta, Armor'stheta, and McDonald'somega coefficients), expected unidimensionality (confirmatory factor analysis), and scalability (nonparametric item response model - Mokken scale analysis). The results indicate that the scale has rather reasonable consistency. Both mean inter-item correlation and corrected mean item-score correlation were relatively high (r= .469 and r = .636 respectively). Additionally, all estimated consistency coefficients reached required thresholds (namely: ? = .88,ß = .79,? = .86,? =.88). Robust confirmatory factor analysis and Cronbach-Mesbah curve convergently supported the hypothesized unidimensional factor solution (CFA fit indexes: ?2 (28)= 26.73, p = .143, CFI = .994, TLI = .992, RMSEA = .041, SRMR = .055.). Moreover, Mokken scale analysis indicated that the scale is scalable (scale’s H = .496) and satisfies the criteria of both monotone homogenity model and double monotonicity model (no significant violations were present). Consistency indices, confirmatory factor analysis, and Mokken scale analysis consistently suggest that the scale assesses a unidimensional construct with reasonable reliability.They also indicated that broader scope of worries that may be present under evaluating conditions (mapped by eight items) tend to occur simultaneously, plausibly without any finergrained structure. The nonparametric item response model suggested that the items allow ranking persons in the same order on the latent continuum and that the ordering of