401-6222-00L  Nonlinear and Robust Regression

SemesterSpring Semester 2015
Lecturers
Periodicitytwo-yearly recurring course
CourseDoes not take place this semester.
Language of instructionEnglish
CommentSpecial Students "University of Zurich (UZH)" in the Master Program in Biostatistics at UZH cannot register for this course unit electronically. Forward the lecturer's written permission to attend to the Registrar's Office. Alternatively, the lecturer may also send an email directly to Link. The Registrar's Office will then register you for the course.


AbstractThe basic ideas of robust fitting techniques are explained theoretically and practically using regression models and explorative multivariate analysis. The second part addresses the challenges of fitting nonlinear regression functions and finding reliable confidence intervals.
ObjectiveParticipants are familiar with common robust fitting methods for the linear regression models as well as for exploratory multivariate analysis and are able to assess their suitability for the data at hand. They know the challenges that arise in fitting of nonlinear regression functions, and know the difference between classical and profile based methods to determine confidence intervals. They can apply the discussed methods in practise by using statistics software.
ContentRobust fitting: influence function, breakdown point, regression M-estimation, regression MM-estimation, robust inference, covariance estimation with high breakdown point, application in principal component analysis and linear discriminant analysis.

Nonlinear regression: the nonlinear regression model, estimation methods, approximate tests and confidence intervals, estimation methods, profile t plot, profile traces, parameter transformation, prediction and calibration
Lecture notesLecture notes are available
Prerequisites / NoticeIt is a block course on three Mondays in June