Abstract: The Present paper reports on the introduction of Risk State Analysis,Automatic clustering of the exposed levels and sieving of the variables are done by the information and non-parametric optimum-seeking criteria proposed by the authors. The risk states are directly analysed overcoming the difficulties encountered in the Selection of the models and multicollinearity of the Logistic and Cox regressloa analysis. This give hi沙er clarity in the analysis results thus helping to show the joint effects of several risk factors. To illustrate the application of this new method, two numerical examples is the case-control study and survival analysis are gwen.