In epidemiological exposure studies frequently one unexposed group and k 1 ex-posed groups are assumed. The number of diseased and heathy persons in every exper-imental group establishes a 2×k contingency table. Then, we can study the outcomeunder different exposure levels. In this paper, several different models are deduced whenthe margins are given in a 2×k contingency table. However, performing tests are equiv-alent in different situations, We are to select one to study test. we propose a consistentmodel selection method to identify the dose response. Further, we demonstrate the per-formance of our method through simulate study. |