Incomplete Multivariate Designs, Optimal With Respect to Fisher's Information Matrix

摘要

In most experiments or investigations, usually more than one response or characteristic is measured on an experimental unit or individual. The interest very often lies in estimating certain parameters (usually of the location or scale type), on each response. It is well known that, in general, as the number of units on which a response is observed is increased, the efficiency of the estimate also increases. However, in general, one is faced with financial limitations. In the paper, the authors consider a multiresponse design problem when such parameters are to be estimated. The optimization under the given cost restriction, is done with respect to maximizing the determinant of the fixed information matrix corresponding to these parameters. (Author)