Multi-Unit Longitudinal Models with Random Coefficients and Patterned Correlation Structure: Modelling Issues

Research project P 14305 Multi-unit Longitudinal Models: Modelling Issues Johannes LEDOLTER 26.6.2000 We propose to study modelling issues in multi-unit longitudinal models with random coefficients and patterned correlation structure. In particular, we plan to study the * Development of improved specification procedures for multi-unit repeated-measures data. * Graphical techniques for model specification and model diagnostic checking * Inference procedures that incorporate order-restrictions among the fixed effects * Estimation of local continuous-time trend components using polynomial smoothing splines * Application of the newly developed procedures to the analysis of several data sets In the following research proposal we describe the planned research activities, we review the relevant literature, and we present several initial research findings. The class of models which is studied, multi-unit longitudinal models, combine both cross-sectional and longitudinal aspects. Many empirical investigations involve the analysis of data structures that are both cross- sectional (that is, observations on several units at a specific time period or a specific location) and longitudinal (that is, observations taken over time or space). Multi-unit longitudinal data structures arise in economics and business where panels of units are studied over time, in biostatistics where groups of patients are exposed to different treatments and followed over time, and in the life-sciences where the response of animals to different treatments is investigated over time.