Confirmatory factor analysis of the AEQ-A questionnaire in Finland

The Adolescent version of the Alcohol Expectancy Questionnaire (Brown et al., 1980) was administered to a male population sample in Finland (N = 195; median age 18.5 years). Three issues were investigated: Did the items of each subscale measure single latent constructs? Was the AEQ-A better described by a single-factor model (e.g., global positive expectancy) or by a null model implying scale independence? Were alcohol expectancies related to drinking habits? Confirmatory factor analyses by PRELIS2 and LISREL8 indicated that single-construct models were appropriate only for the Enhanced sexuality and Increased arousal scales while several items were unrelated to the postulated latent constructs in the other five scales. Adequate fits were obtained for these scales only when unrelated items were removed and/or analyses were based on nondichotomus scores derived from contentwise homogenous groups of items. Even when revised so that the each scale formally fitted a single latent construct, the resulting latent constructs were not well described by either the single-factor model or the null model. Expected social benefit in terms of the revised Enhanced or impeded social behavior scale was the most significant predictor of drinking frequency while quantity consumed per drinking occasion was significantly related to the Improved cognitive and motor abilities scale only.