RASCH ANALYSIS OF THE EATING ATTITUDES TEST: AN EXPLORATORY AND CONFIRMATORY STUDY OF RATING SCALE CATEGORY EFFECTIVENESS, AND CALIBRATION OF A NEW RATING SCALE

Abstract

Proper screening for the risk of having an eating disorder is critical to provide necessary treatment and avoid negative health outcomes. The Eating Attitudes Test (EAT) is one of the most prominent assessment tools for eating disorder risk and is administered in both clinical and non-clinical settings. The EAT uses a Likert scale and relies partially on a total summed score of answers to determine whether a clinical referral for diagnosis and/or treatment is required. Historically, measurement methodologies have neglected to consider that the ordinal nature of total summed scores, or counts, does not provide adequate statistical analysis. In this study, a more advanced measurement model called Rasch, specifically the Rasch Rating Scale Model (RSM) and Rasch calibration, was utilized to analyze the properties of the EAT and to further investigate this valuable tool. The EAT was administered to a sample from a southern college community and an exploratory study was performed to evaluate its category function. A confirmatory study was then performed to substantiate findings from study one. Finally, a calibration was performed to study item difficulty and person ability.

Results show that the EAT with 26 items (EAT-26) may not function as intended. Post-hoc analysis suggested a four or three category Likert scale may be more optimal than the original with six. A confirmatory study provides more evidence including ordered thresholds and satisfactory person separation index and reliabilities. This implies that the participants are better able to distinguish between the categories of the adjusted Likert scale. The calibration of this four category scale provides additional insight into the hierarchy of EAT items and recommendations for future study.