A stepwise clustering algorithm, a method of multivariate statistical analysis, is suggested in this paper. The algorithm is designed for solving problems connected with stepwise regression. It is efficient not only in handling both continuous and discrete variables, but also in the nonlinear relationships between the variables. The above procedure was used in an attempt to find out the causal association of esophageal cancer with its precursors, i.e. nitrates and nitrites of nitrosamines, some of which are known to be carcinogenic. An analysis has been made of the correlation between esophageal cancer as well as severe epithelial hyperplasia of the esophagus and the concentrations of NO3- and NO2- in the drinking water. The samples used were collected from 495 wells in 49 production brigades of the Yaocun Commune in Linxian County, Honan Province. The result indicates that esophageal cancer is definitely connected with the levels of NO3- (summer) and NO2- (spring) in the drinking water. Severe epithelial hyperplasia is defintely connected with the contents of NO2- and NO3- in the drinking water collected in spring, autumn and winter. Our preliminary analysis shows that the stepwise clustering algorithm is a useful statistical method to be used for medical research.