| Abstract: | This study develops simultaneous tests for shifts in marginal distributions pertaining to complete two-way multivariate data not having Gaussian errors. Gerig's (1969) statistics based on ranks are used, a special case yielding Friedman's (1937) χ Statistics statistics for the one-dimensional marginal distributions. Exact α-level procedures are not now available. Instead, approximate but conservative tests are constructed using available tables together with suitable probability inequalities. A variational result permits unlimited tests even when these are suggested by the data. The procedures are illustrated using data from field studies with peanuts; some extensions are noted; and selected percentage points are provided to support some of the tests. |
