Modelling individual differences in change patterns
09 / 2006 - unknown
Nederlandse Organisatie voor Wetenschappelijk Onderzoek - NWO
The aim of the current project is to provide more insight into the way people change, by modelling repeated categorical variables. Examples of application include: 1. Neuroscience, where patients can be classified into different stages of typical Alzheimer and three classes of atypical Alzheimer s disease (Lambon Ralph et al., 2003). 2. Genetics, where patients having different types of cancer can be classified based on DNA expressions (Golub et al., 1999). Early classification is assumed to have a major impact on the success of treatment of the disease later on, which emphasizes the need for early repeated measurements of people at risk. 3. Criminology, where recidivism data are often collected (Bijleveld and Mooijaart, 2003). 4. Educational psychology, where children are measured before and after instruction in, for example, mathematics (Van Putten, Van den Brom-Snijders and Beishuizen, 2005). 5. Developmental psychology, where children are observed during a therapy session (Harinck and Hellendoorn, 1987). 6. Political science, where it is of interest whether people change their vote (Rosema, 2004). In all these cases assessing whether change occurred and describing the nature of change is of utmost importance. For categorical longitudinal data this is problematic due to the number of parameters usually involved and the difficulty of parameter interpretation (Molenberghs and Verbeke, 2005). To overcome these problems, De Rooij and Heiser (2003, 2005) and De Rooij (2001a, 2001b, 2002, 2005a, 2005b, 2006) developed distance-association models, that reduce the number of parameters and simplify the interpretation by making a map with the categories of the variable on basis of which people make their moves. The present research has two major aims: 1) Extending the current distance-association models by using covariates, random effects, and mixtures to model individual differences; and 2) solving two important statistical issues: stability of distance-association models and selection of a suitable model.