A test of simultaneous homogeneity of main effects of several factors against an alternative hypothesis with simple order restrictions in main effects of more than one factor in a multifactorial design is considered. This can be regarded as an extension of Shoracks (1967) work where the alternative hypothesis involves simple order restriction in main effects of one factor. We derive the likelihood ratio test statistic, E-2, and its null hypothesis distribution which involves the convolutions of probabilities P(l, k) used in the statistical inference under order restriction. The powers of the E-2 test and the usual F test are compared by simulation.
All Science Journal Classification (ASJC) codes
- Statistics and Probability