关于Cox模型的Bootstrap区间估计和假设检验

目的 提出Ｃｏｘ模型参数的Ｂｏｏｔｓｔｒａｐ区间估计和假设检验。方法 通过对样本进行Ｂｏｓｓｔｓｔｒａｐ抽样，构造Ｂｏｓｓｔｓｔｒａｐ区间估计和假设检验３区域。结果 实例分析表明，当再抽样容量较大时，Ｂｏｏｔｓｔｒａｐ区间估计和假设检验与直接对原样本应用正态性得到的结果相差不大，前者在后者周围波动，假检验结果也趋于一致。结论 Ｂｏｓｓｔｓｔｒａｐ区间估计和假设检验在一定程度上可以作为按正态性估计误

On Bootstrap Confidence Interval and Hypothesis Test of Cox Model

Objective To put forward Bootstrap confidence interval and hypothesis test of Cox model.Methods By resampling from the primary sample to get bootstrap samples,then set up bootstrap confidence interval and hypothesis test.Results When sample is enough,the example shows that confidence intervals and hypothesis tests which set up according to bootstrap method and normality respectively are of asymptotic consistency.Conclusion On certain degree,bootstrap confidence interval and hypothesis test can replace results made from normality,which is more robust when sample is not enough.