Inference for the model of several treatments and a control

Cohen et al. (2000, J. Multivariate Anal., 72, 50-77) noted that for the tree-order model, under exponential family distribution theory, likelihood inference may have shortcomings. In light of this we offer alternative inference procedures for this model under normal distribution theory and in the nonparametric case as well. We suggest new estimators that are not only free of the practical shortcomings of maximum likelihood estimators, but that have other desirable properties as well. We suggest an alternative to the likelihood ratio test which has desirable practical properties and competitive power properties. In the nonparametric case we offer an alternative to the popular tests based on ranks. (C) 2002 Elsevier Science B.V. All rights reserved.