偏正态单向分类随机效应模型下暴露水平的Bootstrap推断

Bootstrap Inference of Exposure Level with Skew-normal One-way Classification Random Effect Model

为评估工作环境中的暴露水平,基于偏正态单向分类随机效应模型,研究暴露水平的区间估计和假设检验问题。首先,利用EM算法给出未知参数的极大似然估计。进而,基于Bootstrap方法,构造个体平均暴露水平的三种Bootstrap置信区间。Monte Carlo模拟结果表明,修正的Bootstrap百分位置信区间在覆盖概率意义下优于其他两种Bootstrap置信区间,Bootstrap标准置信区间在置信上限意义下优于其他两种Bootstrap置信区间。最后,将上述方法应用于苯乙烯暴露数据的案例分析,以验证所提出方法的有效性和合理性。

To assess exposure level in a work environment,we consider the interval estimation and hypothesis testing problems of exposure level based on the skew-normal one-way classification random effect model.Firstly,the EM algorithm is used to give the maximum likelihood estimation of unknown parameters.Secondly,based on the Bootstrap approach,three types of Bootstrap confidence intervals for the individual average exposure level are constructed.The Monte Carlo simulation results indicate that the improved percentile Bootstrap confidence interval performs best in the sense of coverage probability,and the Bootstrap standard confidence interval performs best in the sense of upper confidence limit.Finally,the above approaches are applied to the real data example of styrene exposures to verify the reasonableness and effectiveness of the proposed approaches.