Asymptotic absolute efficiency of projection tests with an application to n-ranking tests.

*(English)*Zbl 0555.62039This paper is concerned with the comparison of tests for different treatment effects in randomized block designs which are based on rankings within the blocks. First some asymptotic properties, for example strong law of large numbers and asymptotic normality under closed alternatives of so-called projection tests are derived.

Next it is shown that Friedman and Page type tests are uniformly efficient and Anderson’s test is at least locally efficient within a large class of n-ranking tests. A direct comparison between the efficiencies of the Anderson and Friedman tests is given.

Reviewer: R.Mnatsakanov