轉換點(change point)問題出現在許多的統計領域，包括了有母數、無母數、迴歸、時間序列、序貫、貝式等模型，本研究主要是針對簡單迴歸模型的估計單一轉換點之問題作探討，若我們均勻分布解釋變數並採用最小平方法來估計轉換點，模擬結果告訴我們估計值會有雙峰的現象，此現象造成了變異數的增大。我們嘗試利用二階段設計來改善之前的估計，藉由第一階段所得到轉換點的可能範圍來估計第二階段的實驗，模擬結果顯示此兩階段估計的確降低了估計值的差異。 / The change point problem can be involved in many models such as parametric, nonparametric, regression, time series, sequential, Bayesian, and so on. This thesis focuses on estimating the location of the change point in a simple regression model. We first show that the computational simulation demonstrates a bimodal phenomenon which could increase the variation of the estimator badly when the independent variables are allocated uniformly over the explanatory interval and the least square method is used to determine the estimator of the change point. In the second part,we implement a two-stage design that tries to shrink the possible location of the change point via first stage and then design the second stage accordingly. Simulation result gives a positive response in reducing the variation caused by the bimodal phenomenon.