Rasch analysis of measurement instruments capturing psychological personal factors in persons with spinal cord injury
Claudio Peter, Stefan E. Schulenberg, Erin M. Buchanan, Birgit Prodinger, Szilvia Geyh
Participation & Social Integration Unit, Swiss Paraplegic Research, CH-6207 Nottwil, Switzerland. E-mail: firstname.lastname@example.org
Objective: To evaluate the metric properties of distinct measures of psychological personal factors comprising feelings, beliefs, motives, and patterns of experience and behaviour assessed in the Swiss Spinal Cord Injury Cohort Study (SwiSCI), using Rasch methodology.
Methods: SwiSCI Pathway 2 is a community-based, nationwide, cross-sectional survey for persons with spinal cord injury (SCI) (n = 511). The Rasch partial credit model was used for each subscale of the Positive Affect Negative Affect Scale (PANAS), Appraisal of Life Events Scale (ALE), Purpose in Life test – Short Form (PIL-SF), and the Big Five Inventory-K (BFI-K).
Results: The measures were unidimensional, with the exception of the positive affect items of the PANAS, where pairwise t-tests resulted in 10% significant cases, indicating multidimensionality. The BFI-K subscale agreeableness revealed low reliability (0.53). Other reliability estimates ranged between 0.61 and 0.89. Ceiling and floor effects were found for most measures. SCI-related differential item functioning (DIF) was rarely found. Language DIF was identified for several items of the BFI-K, PANAS and the ALE, but not for the PIL-SF.
Conclusion: A majority of the measures satisfy the assumptions of the Rasch model, including unidimensionality. Invariance across language versions still represents a major challenge.