The Theory and Applicatoins of Generalized Logistic Regression for Risk Analysis of Fuzzy States(2)—The Methods of Fitting the “Best” Non-Saturated Model and Estimating Attributable Risk

By the aid of fuzzy state analysis,the Cross-Product Differance Sum and the Akaike's Information Criterion,The metho ds, fitting the best non-saturated model and estimating attributable risk,are suggested in this papper.The realization of fitting the best non-saturated model,breaking through some constraints of nested hierarchy models,can improve the abilities exploring joint actions of weakly correlative-influencing factors.