Interpretation of conditional and marginal regression models for correlated non-normal data

Date

Dec 11, 2014

Time

2:30 PM - 3:30 PM

Speaker

Dheban Srirangan

Affiliation

IMB doctoral candidate

Series

IMB - Seminar

Language

en

Main Topic

Mathematik

Other Topics

Informatik

Host

IMB

Description

GLMMs represent the class of the most general regression models. These models are able to incorporate both fixed and random effects and generalize LMMs to distributions of responses that belong to the exponential family. However, in case of non normal responses the fixed effects can no longer be interpreted on a population level (marginal level). In contrast, GEEs (Generalized Estimation Equations) allow such an interpretation deriving from the usage of a quasi-likelihood as well as an assumption on the correlation of responses within a subject/cluster. A comparison of both model types regarding the interpretation of regression coefficients will be addressed.

Last modified: Nov 26, 2014, 12:29:56 PM