Measurement and prediction in culturally heterogeneous population

Annette Maij-De Meij PhD

Project: Department of Work and Organizational Psychology, VU University Amsterdam

Project financed by: The Netherlands Organisation for Scientific Research (NWO)

Supervisors: Prof. dr H. Kelderman, prof. dr H. van Flier

Project running from: 1 November 2002 – 1 November 2006

Summary:
Psychological measurement and prediction procedures usually assume that individual differences can be described by one standard measurement and prediction model. However, empirical psychometric research has shown that these models are not always invariant over cultural groups. Therefore, it is important for both comparative and applied psychology to study these qualitative differences in measurement and prediction and to design culture-fair procedures that take these differences into account. For example, in selection psychology, candidates from different cultural groups have to be compared with respect to their suitability for a certain job. In practice, information about group membership is often used in the decision. This practice has been criticized, because it may result in different selection probabilities for equally fit individuals from different groups. It is expected that this problem will increase as minority groups become more heterogeneous as a result of differences in acculturation. We propose new measurement and prediction methods based on mixture measurement models. The methods take qualitative group differences in test performance into account, but do not rely on information about group membership. Selection decisions are solely based on item response patterns. Qualitative group differences are only taken into account to the extent that they may be determined from the response patterns. Individual scoring rules are derived on the basis of a moderated directed acyclic graphical model containing a mixture measurement model that describes the cultural impact on measurement and prediction. Statistical, substantive, and applied issues are addressed. Available and newly collected experimental data will be analyzed. The ultimate aim of the project is to improve the validity of test scores of immigrants.

Date of defence: 13 June 2008

Title of thesis: Measurement and prediction in heterogeneous populations

ISBN: 978-90-86592-03-6