Multiple Group Time-Series Design

Abstract

Configurations of means denying treatment effectiveness are presented for a multiple group time-series design. The configurations imply a sequence of null hypotheses. Failure to reject all of these hypotheses means the treatment effect is not supported. Rejection of any one of the hypotheses leads to tests of more detailed hypotheses. The rejection of these hypotheses can provide support for a treatment effect. Since both univariate and multivariate test criteria can be used to test the hypotheses, the procedure can be applied to data sets that have fewer subjects than occasions of measurement.