School Seminars and Colloquia

Penalized Spline Regression and Mixed Models - a Promising Alliance

Statistics Seminar

by Goeran Kauermann


Institution: School of Economics and Management, University of Bielefeld, Germany
Date: Tue 18th November 2008
Time: 1:15 PM
Location: Room 213, Richard Berry Building, The University of Melbourne

Abstract: Penalized spline fitting as smoothing method has achieved recognizable popularity over the last years. Tracing back to Eilers und Marx (Statistical Science, 1996) the book by Ruppert, Wand & Carroll (2003, Cambridge University Press) shows the versatility and flexibility of the approach.
The presentation starts with an introduction to penalized spline smoothing. Basic ideas are illuminated and supplemented by data examples. In particular the link between penalized spline smoothing and linear mixed models will be exhibited. This allows to select the smoothing parameters based on the maximum likelihood principle. The idea is extended with a number of research results including smoothing in the presence of correlated errors, local adaptive smoothing and duration time modeling. Each extension is motivated and supported with a data example.

For More Information: Dr. Guoqi Qian, g.qian@ms.unimelb.edu.au