负二项分布抽样中的患病率无偏估计

目的本次研究以第三次全国血吸虫病流行病学调查为背景,对其部分抽样过程进行计算机模拟,采用负二项分布抽样方法,得到感染率的无偏估计,并与传统的抽样方法比较,综合评价两种抽样方法的优缺点。方法分别在样本量相同及样本量不同两种情况下对抽样结果估计感染率的绝对误差、相对误差及正确率作统计学描述分析,并综合评价。结果在相同样本量下,两种抽样方法估计的感染率在绝对误差、相对误差、正确率及可信区间宽度方面差别的P值均大于0.05(当感染率为0.6%时,两者的正确率及可信区间宽度差别P值接近0.05);在样本量不同时,两种抽样方法估计的感染率在正确率方面差异无统计学意义(P值均大于0.05),在绝对误差、相对误差及可信区间方面差别的P值均小于0.01,仅在感染率较高时(大于10%)两者差异无统计学意义。结论在样本量一致情况下,两种抽样方法在不同的感染率范围内的估计精度相当。当实际感染率较小时(如小于1%),采用负二项分布抽样可实现抽到足够的患者;当实际感染率未知且无法预测时,该方法又是一种探索性的抽样方法。

Unbiased Estimation Method of the Prevalence by Negative Binomial distribution Sampling

Objective To obtain an unbiased estimation of the prevalence of schistosomiasis by negative binomial distribution sampling(NBS) and assess the advantages and disadvantages of traditional sampling method(TS) and negative binomial distribution sampling method.Methods To derive the unbiased estimation and variance of the prevalence of schistosomiasis by NBS,random simulation is performed to assess the advantages and disadvantages of the two sampling methods.Results For the same sampling size,there are no significant differences between the two methods in the absolute error,the relative error,the correctness rate and the average width of CI (when the prevalence is less than 0.6%,the P values of the correctness rate and the average width of CI are very close to 0.05);for different sampling size,there are significant differences between the two methods in the absolute error,the relative error and the average width of CI,but no significant differences in the correctness rate.When the prevalence is larger than 0.1,the results show no significant differences.Conclusion Under the same sampling size,those two sampling methods have the same accuracy.When the actual prevalence is very low (1% e.g.),NBS is superior to TS.Because it can make sure to get enough patients.When the prevalence is unknown and unpredictable,NBS is also an exploratory method.