Listy Biometryczne - Biometrical Letters vol. 42(1), 2005, pp. 43-56
|
A-, D- AND E-OPTIMAL DESIGNS FOR QUADRATIC AND CUBIC GROWTH CURVE MODELS WITH CORRELATED ERRORS
Katarzyna Filipiak, Anna Szczepańska Department of Mathematical and Statistical Methods, Agricultural University of Poznań, Wojska Polskiego 28, 60-637 Poznań, e-mail: kasfil@au.poznan.pl, sanna6@wp.pl |
In this paper polynomial growth curve models with correlated errors are considered. We study A-, D- and E-optimal designs for quadratic and cubic growth curve models with three and four time points on the time interval [0,2]. Optimal designs for four correlation structures of errors are compared: nearest-neighbor, autoregression, moving average and unstructured antedependence.
Growth curve model, information matrix, covariance structure, optimal design.
