[Research strategies in the evaluation of depressive states]. / Stratégies de recherche dans l'évaluation des états dépressifs.

The present paper discusses the respective benefits of a latent trait model and of factorial correspondence analysis in the research strategies of depression inventory construction. Results on the processing of the same list of depressive symptoms by both analyses are presented. The fact that, in questionnaires based on latent models, each answer brings an information closely dependent on its place in the hierarchical order is underlined. The latent trait strategy construction offers an accurate way to assess how patients recover from depression.

Measuring the severity of depression through a self-report inventory. A comparison of logistic, factorial and implicit models.

The present study investigates the psychometric adequacy of a self-report inventory of depression (the QD2) with respect to the severity of depressive symptoms, within the framework of a unidimensional model of depressive states. Responses to the 52 QD2 items, by 481 subjects divided into four groups (depressed, psychiatric, somatic and control), were subjected to a Rasch analysis. Results show that 15 items (QSD(R1] display a satisfactory goodness of fit for both items and persons, as defined by the Rasch model. This set of items, solely related to 'pure' depression, gives a measure of the severity of depression, free from anxiety symptoms. The discussion focuses upon the advantages of the Rasch model as compared to traditional tests. The Rasch 15-item solution should provide a parsimonious tool for future studies on the classification of depressive disorders.