Model testing, model fitting, and social decision schemes.

Schönemann (1979) suggests that Davis' (1973) social decision scheme (SDS) model in general and Davis et al.'s (1977) application of that model in particular suffer from a model identifiability problem. It is argued here that both the general model and its application in Davis et al. (1977) deal adequately with this problem. The basic flaw in Schönemann's reasoning is traced to a failure to recognize the different ways in which the SDS model has been used. We distinguish two common uses of the model: one which can adequately deal with the identifiability problem and the other for which the identifiability problem is not relevant. After discussing these different ways of applying the SDS models, we argue that Schönemann's conclusions are, for the most part, unjustified.