Abstract
The Anorexia Nervosa Inventory for Self-Rating (ANIS) was the first instrument covering aspects of the general psychopathology of eating disorders alongside eating attitudes, but its factorial integrity in non-clinical samples has not yet been investigated. Thus, this report is aimed at assessing the factorial structure of the ANIS with different methods of extraction and rotation. Data from a population-based random sample of German-speaking Italian schoolgirls aged 11-20 years (n = 1402) were used. The instruments included the ANIS and the Eating Behaviour Severity Scale. The internal consistency (Cronbach's alpha of the ANIS and four of its subscales was 0.70-0.88; however, reliability of the subscales "Sexual Anxieties" and "Obsessive-Compulsive Traits" was doubtful in non-clinical female adolescents. The original six factors were convincingly replicated by maximum likelihood extraction and principal component analysis. As some items and subscales showed weaknesses, a shortened 20-item version was derived containing the subscales "Figure Consciousness", "Feelings of Inadequacy", "Adverse Effects of Meals", and "Bulimia". All items showed factor loadings > 0.50 and item-total-correlations between 0.30 and 0.80. Internal consistency of the ANIS-20 total scale (0.90) was remarkable as it was for the subscales (0.70-0.85); discriminant validity as assessed by a cross-validational approach (random split-half samples) was not affected by this abbreviation. In conclusion, both the original ANIS and the ANIS-20 are psychometrically sound instruments. Because brevity is important in screening studies, the short form is recommended.