Lecture on Generalized Regression
Prof. Gerhard Tutz and Prof. Helmut Küchenhoff
02.02.2012
Prof. Tutz and Prof. Küchenhoff gave a lecture on February 1, 2012 about General Regression Analysis.
The course introduced into modern ways of structured regression. The regression problem is split into two components, the structural component, which links the explanatory variables to the expected response and the stochastic component, which specifies the noise.The splitting into two components allows modifying both components separately, which yields to generalized linear models. Generalized linear models are a unifying framework that allows for different distributions of the response, including, among others, normal, binary and Poisson distributions. The specification of a link function allows letting the response depend on explanatory variable in a more flexible non-linear way. The concept is illustrated by various applications.
The course introduced into modern ways of structured regression. The regression problem is split into two components, the structural component, which links the explanatory variables to the expected response and the stochastic component, which specifies the noise.The splitting into two components allows modifying both components separately, which yields to generalized linear models. Generalized linear models are a unifying framework that allows for different distributions of the response, including, among others, normal, binary and Poisson distributions. The specification of a link function allows letting the response depend on explanatory variable in a more flexible non-linear way. The concept is illustrated by various applications.