Statistiktage 2007 - Kongress der Österreichischen Statistischen Gesellschaft ÖSG
Sprache des Tagungstitel:
Deutsch
Original Kurzfassung:
We will discuss the efficient designs for the parameters of correlated processes. We provide a study of small sample and asymptotical comparisons of the efficiencies of equidistant designs with taking into account both the parameters of trend as well as the parameters of covariance. If only trend parameters are of interest, the designs covering more-less uniformly the whole design space are rather efficient when correlation decreases exponentially. The role of the nugget effect in D-optimal designs will be explained. Some issues on designing for spatial processes will be also provided. The real data illustrating examples will be provided: D-optimal designs helping to recovery Dark Age and some examples in context of spatiotemporal modeling. Keywords: D-optimal design, efficiency, equidistant design, correlated process.