Search

Bias and mean squared error for linear combinations of the isotonic regression estimators are computed. The case of sampling three distinct populations and the case of sampling seven or fewer populations having common mean are studied in detail. Numerical results are given, and comparisons between isotonic and unbiased estimation procedures are made. Le biais et l'erreur quadratique moyenne des estimateurs isotoniques sont calcules. On etudie en detail le cas de trois populations distinctes, et celui de sept populations ou moins de meme moyenne. On enonce des resultats numeriques et on etablit une comparaison entre l'estimation isotonique et l'estimation sans biais.

The exact distribution of a nonparametric test statistic for ordered alternatives, the so-called $\bar{\chi}^2_{\operatorname{rank}}$ statistic, is discussed. Two tables are given.