The background to loglinear modeling is described. Loglinear modeling permits researchers to try different models in the same way that multiple linear regression is used to fit data. This is useful inexploratory analysis. The article considers how the effects of eachof the many possible variables and the relationships among them aresegregated. In this study, the multiway contingency table approach was used. This approach is based on the principle of minimum discrimination information (mdi), which leads to exponential families of the parametgers yielding loglinear models. A key feature of the mdi approach is that it provides the necessary rationale for a uniform treatment of the many ad hoc procedures and seemingly different approaches to contingency table analysis. estimation problems are discussed.
Abstract