Improving psychometric assessment of the Beck Depression Inventory using multidimensional item response theory.

We studied the latent factor structure of the Beck Depression Inventory (BDI) under the light of Multidimensional Item Response Theory models. Under a Bayesian Markov chain Monte Carlo setting, we chose the most adequate model, estimated its parameters and verified its fit to the data. An evaluation of the inventory in terms of the assumed dimensions seems to agree with previous investigations in the factor structure of the BDI present in the literature. Cognitive and somatic-affective latent traits were identified in the analysis making possible the interpretation of symptom evolution along these dimensions, in terms of probability of their appearance.