论文部分内容阅读
目的应用Monte-Carlo模拟进行基于人时的相对危险度的分布估计。方法结合实例进行相对危险度的模型构建、拉丁超立方抽样和概率分布的拟合及RR可信区间的几种计算方法比较。结果模拟的RR频率分布经拟合符合Pearson5、Lognorm、Gamma和InvGauss4种分布,以Pearson5分布拟合最佳。模拟的RR值95%可信区间结果与统计量函数计算值、Wald法和Score法大致相当,但其上限值和下限值均略小。结论应用Monte-Carlo模拟结合拉丁超立方抽样技术,实现了基于人时的相对危险度的分布估计,该方法可应用于更为复杂的参数分布估计。
Objective To use Monte-Carlo simulation to estimate the distribution of relative risk based on the time of the person. The method is combined with examples to build relative risk models, Latin hypercube sampling and fitting of probability distributions, and several calculation methods for RR confidence intervals. Results The simulated RR frequency distribution was fit to Pearson5, Lognorm, Gamma, and InvGauss. The Pearson5 distribution was the best fit. The simulated 95% confidence interval (RR) of the RR value is roughly equivalent to the calculated value of the statistical function, the Wald method, and the Score method, but the upper and lower limit values are slightly smaller. Conclusion The Monte-Carlo simulation combined with the Latin hypercube sampling technique realizes the estimation of the relative risk based on the man-hour, and this method can be applied to more complex parameter distribution estimation.