Chenhcheng Hao: Explicit influence analysis for count data under AB–BA crossover studies
Time: Wed 2014-10-01 13.00 - 14.00
Location: Room B705, Department of statistics, Stockholm university
Participating: Chenhcheng Hao, Stockholm university
Generalized mixed linear models (GMLMs) are common utilised in the crossover studies with count data. The authors describe a method to assess influence of subjects on the maximum likelihood estimates under the GMLM-type crossover models. Various perturbations are performed on the observed data. By decomposing the GMLM-type crossover models, the authors achieve explicit expressions of the influence on the estimated treatment effect. The influence measure turns to relate to the Pearson residuals of the subjects in the unperturbed model.