"statistische Modellbildung" Zu Diesem Kolloquium Wird Eingeladen 4. Herbstkolloquium Des Graduiertenkollegs "statistische Modellbildung" Design-modifications with Control of the Type I Error Rate: Extensions of the Crp-approach Considering Nuisance Parameters or Multiple Selection 16.30 H Pierre Dr

Optimal designs in multivariate linear models Diskussionen zu den Projektbereichen For chronic diseases, cross-over designs are the most common experimental devices for comparing treatments and selecting the most efficient one. During the last decade, the theory of optimal cross-over designs has been considerably developed, especially due to the seminal paper by Kushner (1997). In this talk we present the state of art of optimal cross-over designs and we show that the optimal design may highly depend on the analysis model and on the parameter of interest. Some considerations on estimation of the given parametric function under the general Gauss-Markov model In this talk, we consider estimation of the given estimable parametric function under the general Gauss-Markov model. In particular, we try to highlight some differences which occur between the cases of estimation of the expected value and estimation of the given parametric function. Special emphasis is given to the concepts of linear sufficiency and linear completeness, and to the problem of when the ordinary least squares estimator equals the best linear unbiased estimator.