On the Robust Linear Invariant Prediction of Order Statistics
武司 石川, 孝一 小谷 · Institutional Repositories DataBase (IRDB) · 1987
In this article, robust linear invariant predictors for future order statistics from a population with unknown location, scale and shape parameters are studied. The shape parameter is only known to be a member of a set of several possible values. Simple linear invariant predictors for future order statistics are presented under the above situation. And the linear efficiencies for their predictors are compared with optimally robust linear invariant predictors. The simple linear invariant predictor is found to be quite close to the optimal robust linear invariant predictor.