透過您的圖書館登入
IP:18.118.2.15
  • 期刊

Survival Analysis with Misspecified Models

錯誤設定模型之存活分析探討

摘要


在存活資料分析中,當模型設定錯誤時,大部分的迴歸程序參數估計是非常敏感的。本研究針對一般常用的兩個線性秩統計量(Cox估計量和指數迴歸估計量)與Buckley-James估計量作比較分析,同時探討當存活資料應該用兩個自變數來解釋存活時間時,但在統計分析中卻忽略了其中一個自變數所造成對參數估計的變化。 電腦模擬產生兩種自變數的資料:兩個獨立變數和兩個相關變數;結果顯示,不論資料是否有設限(Censored),Cox估計量均爲有偏差的。指數估計量唯有在資料是沒有設限時才是不偏的,而Buckley-James估計量在所有的模擬研究中,其值均接近不偏。因此,當在省略一個必須存在於模型中自變數的資料分析時,建議可使用Buckley-James估計量來取代Cox估計量或指數估計量。

並列摘要


In survival data analysis, most regression estimators are very sensitive to model misspecification. The paper studies the bias of two linear rank statistic estimators (the Cox estimator and the Exponential regression estimator) and the Buckley-James estimator for data involving two covariates that is analyzed using only one covariate. Simulations are performed when the covariates are independent and when they are correlated. The Cox estimator for the coefficient of the covariate included in the model is shown to be biased regardless of whether there are censored data. The Exponential estimator is unbiased when there is no censoring, but is biased when there is censored data. For all cases studied, the Buckley-James estimator is approximately unbiased. When it is believed that a necessary covariate is being omitted, the Buckley-James estimator is recommended.

參考文獻


Breslow, NE(1972).Discussion' of Cox DR. Regression models and life tables.J. Roy. Statist. Soc. Series B.34,216-217.
Buckley, J.,James, I.(1979).Linear regression with censored data.Biometrika.66,429-436.
Cox, D. R.(1972).Regression Model and Life Tables (with discussion).J. Roy. Statist. Soc. Series B.34,187-220.
Cox, D. R.,Oakes, D.(1984).Analysis of survival data.London:Chapman and Hall.
Heller, G.,Simonoff, J. S.(1990).A comparison of estimators for regression with a censored response variable.Biometrika.77,515-520.

延伸閱讀