Person-misfit in item response models explained by means of nonparametric and multilevel logistic regression models

Judith Conijn
MTO, Tilburg School of Social and Behavioral Sciences
Tilburg University

Project: NWO project at Tilburg University

Project running from: 1 October 2007 – 1 December 2012

Supervisors:
– Prof. dr K. Sijtsma (Tilburg University)
– Dr M.A.L.M. van Assen (Tilburg University)
– Dr W.H.M. Emons (Tilburg University)

Summary:
Performance on psychological tests and personality inventories may be unexpected. This may be due to cheating or test anxiety (achievement testing), or response inconsistency or lack of traitedness (personality). Traditional person-fit measures are primitive in that they only flag unexpected performance but do not provide explanatory information. Two recent approaches provide more explanatory information. One is flexible (i.e., nonparametric) but only suggests an explanation. The other is not as flexible (i.e., parametric) but explicitly uses auxiliary information in a multilevel framework. Both approaches are studied and integrated so as to provide a better understanding of individual test performance.

Date of defence:   27 March 2013

Title of thesis:      Detecting and explaining Person Fit in non-cognitive measurement

ISBN/EAN:   978-90-5335-659-3