The Optimality of Single-group Designs for Certain Mixed Models |
| |
Authors: | Thomas Schmelter |
| |
Institution: | (1) Institute for Mathematical Stochastics, Otto-von-Guericke-University, PF 4120, 39016 Magdeburg, Germany;(2) Schering AG, Berlin, Germany |
| |
Abstract: | In this paper optimal designs for the estimation of the fixed effects (population parameters) in a certain class of mixed
models are investigated. Two classes of designs are compared: the class of single-group designs, where all individuals are
observed under the same approximate design, and the class of more-group designs with the same mean number of observations
per individual as before, where each individual can be observed under a different approximate design. It is shown that any
design that is Φ-optimal in the class of single-group designs is also Φ-optimal in the larger class of more-group designs.
The considered optimality criteria only have to satisfy mild assumptions, which is eg the case for the D-criterion and all
linear criteria. |
| |
Keywords: | Optimal design Mixed model Random coefficient regression Approximate design |
本文献已被 SpringerLink 等数据库收录! |
|