New statistical methods under biased sampling and its application to medical studies

在医学研究的统计分析中，数据抽样往往是有偏差的。在这项研究中，新的统计方法已经开发出这种情况。首先，我们构造了一个可信区间，它对荟萃分析中的发表偏倚具有鲁棒性。然后，对于抗高血压药物的临床试验，我们提出了一种评估治疗与治疗之间因果相互作用的方法，控制混杂因素的影响。最后，在数据由于删失而有偏倚的生存分析中，我们考虑了一种方法，该方法使我们能够解释考克斯回归模型的分析结果，即使它没有被正确指定。

In statistical analysis for medical studies, data sampling is often biased. In this research, new statistical methods have been developed for such situations. Firstly, we constructed a confidence interval which is robust to publication bias in meta-analysis. Then, for clinical trials of antihypertensive drugs, we proposed a method for evaluating causal interaction between the therapy and the treatment, controlling the effect of confounding factors. Finally, in survival analysis where the data is biased due to censoring, we considered a method which enables us to interpret a result of the analysis by the Cox regression model even when it is not correctly specified.