P2-4 Effect Separation in Regression Models with Multiple Scales

PhD student: Hauke Rennies
Thesis Committee: Thomas Kneib, Tatyana Krivobokova, Helmut Herwartz

In regression analyses of spatially structured data, a common practice is to introduce a smooth function of space (or spatially correlated random effects) into the model to account for this information and to reduce or even avoid bias in the estimation of the effects of covariates. If also the covariates are spatially correlated, the spatial random effects may confound the effect of the covariates. In these cases the model fails at identifying the true covariate effect. Similar problems might arise in time series or clustered data.

The aim of this project is to extend existent or to develop new regression techniques to overcome the difficulties of effect separation in regression models with multiple scales.

back to overview of all projects