Handling missing scores and outliers in test and questionnaire data

Joost van Ginkel PhD
Department of Methodology
Faculty of Social Sciences, Tilburg University

Supervisors: Prof. dr K. Sijtsma, dr A. Van der Ark, prof. dr J.K. Vermunt

Project running from: 1 March 2003 – 1 March 2007

Summary
Nowadays, much software is availabe for the analysis of the scores on J items from tests and questionnaires using item response theory (IRT). From the perspective of practical data analysis, a limitation in several of these software packages is that they cannot handle missing data other than by means of list-wise deletion. Although often they are unaware of the damage this may do to their data analysis, this is the option used by most researchers. In addition, many researchers are unfamiliar with the concept of outliers, which may also distort the results of an IRT analysis. The purpose of this project is to find simple and practical solutions for handling missing data and outlier problems in test and questionnaire data, and to implement these solutions into SPSS code and, more specifically, the much used IRT program MSP.

Date of defence: 8 June 2007

Title of thesis: Multiple imputation for incomplete test, questionnaire, and survey data. ISBN: 978-90-5335-119-2.