Goodman's quasi-perfect model and Hauser's model were devised for mobility table analysis. These models may be called quasiloglinear models. Quasi-loglinear models have such advantage as loglinear model has. For a model can be compared with another one in measures of fit (goodness of fit) and simplicity (degree of freedom). However, a quasi-loglinear model can have equivalent models which have the same expected frequencies but different parameters. Equivalent models can not be compared one another because of the same test statistics measured. In this paper, to cope with the problems equivalent models bring about, I reconsider the parameters of quasi-loglinear model in terms of specification of conceptions used in social mobility studies, and present the framework to explore the most valid model.