La vraisemblance de patrons de réponses : étude de la précision des indices d'ajustement des scores individuels, de leurs points critiques et du taux optimal d'aberrance

Authors: St-Onge, Christina
Advisor: Valois, Pierre
Abstract: This doctoral research on Item Response Theory (IRT)-based Person-Fit Statistics (PFS) is comprised of three studies. This research was divided in such a way so we could study two key concepts: the detection rates and the critical values of PFS. In the first and third study, detection rates were studied. The second study focused on the critical values of a PFS. In the first article, we observed that the PFS were more accurate when they were used with parametric estimated ICCs (ML2P and ML3P), and this was independent of the sample size. It seems necessary to verify the model-data fit before carrying out appropriateness assessment with IRT-based PFS. Following the development of a table of critical values, in the second article, the degrees of confidence were calculated for each interval and these results lead us to believe that the critical values were precise. These critical values were tested and it was observed that the type I error rates were conservative but the detection rates observed for .05 and .10 type I error levels were slightly inferior to the detection rates found in the literature. In the third article, we investigated the optimal aberrance phenomenon, i.e., we observed an increase in the detection rate of PFS with an increase in the aberrance rate until a peak was reached and then an increase in the aberrance rate lead to a decrease in the detection rates of PFS. These last results help us to explain a phenomenon that was never previously studied.
Document Type: Thèse de doctorat
Issue Date: 2008
Open Access Date: 13 April 2018
Grantor: Université Laval
Collection:Thèses et mémoires

Files in this item:
24923.pdf587.93 kBAdobe PDFView/Open
All documents in CorpusUL are protected by Copyright Act of Canada.