In view of the multivariate regression analysis, we proposed a stepwise regerssion analysis of the distibution of a hyperfine parameter with non-negative constraints in simulation of the overlapped M?ssbauer spectrum. In this method only those rational subspectra, which satisfy both importance and non-negative constraints according to the information in the studied M?ssbauer spectrum, could be comprised in the simulated result. The present method showed two advantages, i.e., fewer calculations and the optimum simulation result which is rational and has least mean square error.