Modelos de diagnóstico cognitivo: clasificación de atributos, funcionamiento diferencial del ítem y aplicaciones

Résumé

At present, a variety of cognitive diagnosis models (CDMs) that vary in generality (i.e., complexity) have been proposed. As with most psychometric models, parameters of more general models require larger sample size to be calibrated accurately. In the current work, commonly used general and specific cognitive diagnosis models are systematically explored in terms of attribute classification accuracy (ACA) and differential item functioning (DIF). It is also provided a detailed investigation to help researchers and practitioners evaluate conditions where a general or specific model can be more appropiate. Conditions such as item quality, sample size, test length, true model, and number of attributes are considered in a ACA simulation study, whereas factors such as sample size, item quality, DIF size, DIF type, and number of attributes per item are investigated in two DIF simulation studies, in which it is proposed two new indices for DIF detection. In addition to ACA and DIF studies, the present project provides two examples using real data. One of the data sets comes from an application of a scale designed to detect individuals with Asperger Syndrome and the other comes from TIMSS 2007 fourth grade mathematics assessment. Finally, a special purpose software was designed and develop to perform CDMs estimation.