中文摘要：

本文考虑了纵向艾滋病数据模型响应值预测时的最优稳健设计问题。分别研究了两类预测问题：一类是群体响应均值的预测,另一类是个体观察值的预测。给出了模型方差参数空间上的最大最小和贝叶斯稳健设计,多个可选模型下的稳健设计以及预测点不确定时的稳健设计。提出了求各类稳健设计的模拟退火算法。

英文摘要：

In this paper,optimal robust designs for prediction in longitudinal models for AIDs data are considered.Two prediction problems are studied.One is to predict the population mean,the other is to predict the individual observation.The maximin and（or） Bayesian methods are used to derive robust designs against uncertainty of the variance parameters,many possible models or uncertainty of the predicted point.A simulation annealing algorithm is proposed to compute these robust designs.