Linear rank tests for radomazed block design

A class of locally most powerful linear rank tests for no treatment effect for various randomized block designs are presented. Closed form expressions of the test statistic are derived for the complete block design, the balanced incomplete block design and a special design in which there are missing observations on one treatment only. In addition, a test statistic is UCLiVCU LUL LUC unbalanced incomplete block design. The maximin efficiency robust test is given for a randomized block design* This test maximizes the minimum Pitman efficiency over a class of possible error distributions. Multiple comparison procedures are presented