The Sequential Test Of Change Points In A Nonlinear Model

In this thesis,we give two sequential methods to solve the problem of detecting the change points of nonlinear models.The first method is based on the estimating functions.By constructing the score function and using the cusum theory,we employ the score function to detect the change point.This method is fit for many kinds of models.We show the details of this technique for nonlinear models.The second method is based on the least squared method.We extend the results of linear models to the nonlinear models.The asymptotic distribution of the test statistic and the empirical size and power of the test are obtained.In simulation studies,we generalize two nonlinear models with exponential form,the growth model and compartment model.We implement the two methods proposed in this paper and compare their results.