Supporting construct validity of the Evaluation of Daily Activity Questionnaire using Linear Logistic Test Models.

Abstract

Construct validity is commonly assessed by applying statistical methods to data. However, purely empirical methods cannot explain what happens between the attribute and the instrument scores, which is the core of construct validity. Linear Logistic Test Models (LLTMs) can provide such explanation by decomposing item difficulties into a weighted sum of theoretical item properties. In this study, we aim to support construct validity of the Evaluation of Daily Activity Questionnaire (EDAQ) by using item properties accounting for item difficulties. Dichotomized responses to the EDAQ were analyzed with (1) the Rasch model (to estimate item difficulties), and (2) LLTMs (to predict item difficulties). Seven properties of the items were identified and rated in ordinal scales by 39 Occupational Therapists worldwide. Aggregated metric estimates-the weights used to predict item difficulties in LLTMs-were derived from the ratings using seven cumulative link mixed models. Estimated and predicted item difficulties were compared. The Rasch model showed acceptable fit and unidimensionality for a sample of 42 locally independent EDAQ items. The LLTM plus error showed significantly better fit than the LLTM. In the former, three of the seven properties were not significant, and the corresponding model including only the significant properties was used to predict item difficulties; they explained 77.5% of the variance in estimated item difficulties. A satisfactory theoretical explanation of what makes an activity of daily living task more difficult than another has been provided by a LLTM plus error model, therefore supporting construct validity of the EDAQ.