對於模型配適是否恰當,通常作殘差圖來進行判斷,或者利用缺適性檢定(Goodness-of-fit)來檢驗模式。在本篇文章中,考慮採用累積加總的模型檢查方法(Model-checking techniques based on cumulative sum),此方法不僅可以檢驗模型是否假設錯誤或者存在自然變異,同時也可以提供當模型中的解釋變數之函數形式不適當時的解決方向。在取得分析資料方面,則考慮Surveillance Epidemiology and End Result (SEER) (www.seer. cancer.gov)資料庫中腎臟癌、食道癌及胰臟癌的資料,利用羅吉斯迴歸(logistic regression)與卜瓦松迴歸(Poisson regression)進行配適,並針對羅吉斯模型檢驗是否存在假設錯誤再加以修正。最後利用得到的模型探討在不同狀況下之病患死亡率。 Model-checking techniques (Biometrics 2002, Lin et al.) develops a statistical method in order to check misspecification or natural variation of different models. We focus at discussing the cumulative residuals by considering moving sum. We consider kidney, esophageal, and pancreatic cancer which from the Surveillance Epidemiology and End Result (SEER) (www.seer.cancer.gov) database released in 2012. Moreover, we provide the results of the logistic regression, the Poisson regression by this data. Each cancer is compared, both graphically and numerically.