Research Seminar
MultiLevel Clusterwise Regression (MLCR)Eva Vande GaerKU Leuven | |
| Abstract: | In psychology, many research questions pertain to the prediction of some dependent variable on the basis of a single or several independent variables. Such questions are often answered by performing a regression analysis. However, sometimes the relation between the independent variables and the dependent variable differs across observations. Such differences can be modeled by assigning the observations to different groups with a separate regression model being associated to each group. Such a model already exists, and is called clusterwise linear regression (CR, Späth, 1979, 1981). CR is a one-level model, implying that all observations are assumed to be independent from each other. Hence, the method is not applicable to multilevel data where, for instance, observations are nested within persons. Therefore, in this paper we extend the clusterwise regression methodology to multilevel data in terms of a MultiLevel Clusterwise Regression model (MLCR). MLCR differs from CR in that MLCR clusters the observations on the second level. An algorithm to fit the MLCR model to data will be presented along with some simulation results. Finally, we will discuss the relation between MLCR and the linear mixed-effects model with heterogeneity in the random effects as captured by a mixture distribution (Verbeke & Lesaffre, 1996). |
| Date: | Tue Apr 27, 12:15 pm - 1:15 pm |
| Place: | room 02.51 (Department of Psychology, Tiensestraat 102, 3000 Leuven) |