基于MOVER方法的二项抽样下相关差的置信区间构造

Confidence Intervals Construction for Relative Difference Under Binomial Sampling Based on MOVER Method

相关差又叫作相对风险减损,它主要衡量风险因素或治疗因素对于个体的额外影响,在流行病学研究中有着重要的临床意义.文章在独立二项抽样下,分别利用传统的区间估计方法和MOVER方法构造了相关差的九种置信区间.文章提出的MOVER方法的优点在于:借用MOVER方法,可以利用两个独立二项分布比率的置信区间来构造相关差的置信区间.与传统的区间估计方法相比,该方法的置信区间构造过程不需要计算相关差的渐近方差,同时不需要Fisher信息矩阵及其逆矩阵,可使计算大大简化.此外,文章通过蒙特卡罗数据模拟考察了九种区间估计方法在不同参数设置下的表现性能.数据模拟结果表明,MOVER方法与传统方法相比,MOVER方法可以提供更精确的置信区间.最后,文章通过实际数据案例来演示了所提出的九种区间估计方法的实际应用.

Relative difference,also known as relative risk reduction,mainly measures the additional impact of risk factors or treatment factors on individuals,and has important clinical significance in epidemiological research.This paper presents nine interval estimation methods for relative differences using traditional interval estimation method and MOVER method under independent binomial sampling.The advantage of the MOVER method proposed in this paper is that by borrowing the MOVER method,the confidence interval of the relative difference can be constructed using the confidence intervals of two independent binomial distribution ratios.Compared with traditional interval estimation method,this method does not need to calculate the asymptotic variance of the relative difference in the process of constructing confidence intervals,and does not require the Fisher information matrix and its inverse matrix,which greatly simplifies the calculation.In addition,this paper investigates the performance of nine interval estimation methods under different parameter settings through Monte Carlo data simulation.The data simulation results show that the MOVER method can provide a more accurate confidence interval compared to traditional methods.Finally,this paper demonstrates the practical application of the proposed nine interval estimation methods through actual data cases.