病例對照研究中常用羅吉斯迴歸模型來探討疾病與風險因子之間的關係。在分析資料時,如果使用了不合適的模型,會得到錯誤甚至荒謬的結論,因此須對模型進行適合度檢定以評估模型是否合適。由於適合度檢定的檢定力會隨著樣本數的增加而增加從而造成過度拒絕配適模型的問題。為了控制拒絕沒有完美配適但仍可接受的模型的機率,本文提出利用Lai and Liu (2018)和Nattino et al. (2020)兩種程序修正HL 檢定 (1980)及CC 檢定 (2004)。在模擬研究中發現,大部分情況下以Lai and Liu (2018)修正CC 檢定的方法較佳。最後透過兩組實例示範提出的方法。
The logistic regression model is often used to explore the relationship between disease and risk factors in case-control studies. If an inappropriate model is used for data analysis, wrong or absurd conclusions will be obtained. Therefore, it is necessary to assess the goodness-of-fit of the model. However, the power of the goodness of fit test will increase with the sample size, resulting in the problem of overrejecting the hypothesis of perfect fit. In order to control the probability of rejecting a model that does not have a perfect fit but is acceptable, this paper proposes to apply the procedures of Lai and Liu (2018) and Nattino et al. (2020) to modify the HL test (1980) and CC test (2004). In the simulation study, it is found that the method of CC test modified by Lai and Liu (2018) is better in most cases. Finally, the proposed method is demonstrated through two real datasets.