Multiple classification latent class models

Rink Hoekstra PhD

Project at: Department of Psychology, University of Groningen, The Netherlands

Supervisors: Prof. dr H.A.L. Kiers, prof. dr A. Johnson

Project running from: 1 March 2003 – 1 May 2007

Summary:
In behavioral sciences, statistical techniques are often viewed as a necessary evil: indispensable, but also difficult to understand. The inherent variability of human performance and the unavailability of perfectly reliable measurement instruments necessitate the use of statistical techniques. But even the simplest statistical techniques are often misinterpreted or found difficult to understand by behavioral scientists. One might blame this state of affairs entirely on the incorrect use of mathematically correct statistical techniques. On the other hand, one could also reason that researchers make so many mistakes when using statistical techniques because the interaction between user and technique is not functioning optimally. In this project, we specifically consider the possibility that people have rational reasons for making certain mistakes or performing a task differently than they should. As a consequence, an important research question is whether certain statistical tech–niques (or aspects thereof) are too difficult to understand. The purpose of the present project is to lay the foundations for an attempt to adjust statistical tech–niques to the needs of behavioral scientists. We distinguish two types of issues. The first issue is the mismatch between what users expect from statistics and what statistics can deliver. The second is the question whether the statistical techniques are used appropriately in practice.

Date of defence: 8 October 2009
Title of thesis: The use and usability of inferential techniques
ISBN: 978-90-367-4001-2