Incremental trend tests, with application in epidemiological exposure studies.

                    Prof. Ludwig Hothorn
                 Professor of Bioinformatics  
                    University of Hannover
                      Hannover, Germany


Order restricted inference is often used in environmetric studies,
toxicology, clinical trials, and epidemiology. 
In clinical trials, is can be used to identify the highest
effective dose step.  
In exposure studies, it is used to evaluate whether incremental
trends exist.  Examples will be given.

Tests based on pair-wise contrasts for neighbor doses will be discussed
using a multivariate t-distribution with a special correlation structure.

In exposure studies the categories are sometimes formed with rather
arbitrary cut-points.
Grouping of neighbored categories in relation to certain hypotheses will be
discussed.
A maximum test of these contrasts will be explained. 
The advantage of this new approach will be demonstrated using real exposure
data.