In order for evaluation studies of mental health service interventions to be useful for purposes such as resource allocation, benefits, effects, or utilities need to be examined in conjunction with their associated costs. Since such studies can be very complex, their implementation expensive, and the results controversial, it is surprising that so little attention has been paid to methods for statistical analysis that enables meaningful inference. The global aim of this proposal is to develop a statistical framework for describing and comparing costs and benefits among programs. Specifically, 1. we will develop statistical models that characterize programs in terms of the joint distribution of cost and benefits; 2. we will develop statistical methods for contrasting, ranking and selecting programs in terms of their costs and benefits; and 3. we will test these newly developed approaches on several cost-benefit data sets and compare the results with those obtained by the use of present methods. One procedure to compare programs is based on multivariate "admissibility" criteria. Another is to rank order or select the best program based on a preference measure. Statistical procedures will be developed for testing admissibility, for testing equality of programs with respect to preference measures and for ranking and selection for various models of the distribution of costs and benefits.