Project R-1432

A flexible method for modelling non-normal hierarchical data structures (Research)

In this project, we propose a flexible modeling framework for non-Gaussian hierarchical data, simultaneously and flexible addressing (1)clustering (normal random-effects) and (2) overdispersion. In this project we want to formulate a general and flexible model for arbitrary data types, which allows both phenomena together. Apart from model formulation they will focus on several estimation methods in inferential routes. The proposed model will be used in three principal areas, all belonging to the core research lines of CenStat in the field of biostatistics and bioinformatics:(1) incomplete data (2) evaluation of surrogate response in meta-analysis of randomized clinical studies;(3) complex surveys, toxicological tests and other biopharmaceutical experiments and high-dimensional data structures. The methodology will be applied to data from a range of clinical, biopharmaceutical, epidemiological and social-scientific areas.

01 January 2009 - 31 December 2012