3-4 hours of lectures per week.
Lecturer
Jørgen Granfeldt
Contents
The variance component models that are treated in Statistics 1 are examples of mixed models in the sense that they include fixed effects as well as two or more normally distributed error terms. In Statistics 1 we presented a complete solution for models that were restricted by rather severe balance conditions. In many cases it is clear that a valid statistical analysis requires a mixed model, but the data have a different structure than the models of Statistics 1 allow for. In some cases the data are downright messy or unbalanced, in others one needs to model a correlation in the errors, or the assumption of normally distributed errors is clearly wrong. Fortunately mixed model methodology has improved tremendously in recent years and can handle the difficulties mentioned above. This is what we aim to show in this course.
A fair amount of the advancements has been implemented in the SAS system in the MIXED procedure which will be used extensively to illustrate the analyses.
Prerequisites
Statistics 1 and Statistics 2.
Literature
To be announced later.
ECTS-credits
10.
Semester
Autum 2002.