论文部分内容阅读
目的评价三种年龄调整率可信区间估计方法,探索适合江苏省宫颈癌筛查研究中年龄调整患病率可信区间估计的方法。方法以二项分布正态近似法、Gamma分布法及“确切概率法”进行年龄调整率的区间估计,运用统计模拟考察多种率及阳性数情况下三种方法的区间覆盖率及宽度。结果当样本量较小(阳性数较少)时,确切概率法的覆盖率离理论可信度的偏差及区间宽度均优于Gamma分布法,两者的覆盖率均明显优于正态近似法;随着阳性数增多,三法各自的覆盖率偏差及区间宽度均逐渐变小,方法间的差异亦逐渐缩小;当阳性数增至30以上时,确切概率法及正态近似法的覆盖率的偏差皆在±1%以内,此时两者的区间宽度接近;而Gamma分布法的覆盖率偏差若要达到1%以内,则要求总阳性数在100以上。无论样本构成是轻度偏离还是明显偏离总体构成,上述规律皆成立。结论综合考虑区间覆盖率、区间宽度及计算便捷性,建议当总阳性数小于30时,采用确切概率法计算调整率的可信区间;当总阳性数大于等于30时,采用正态近似法。
Objective To assess the confidence interval estimates of three age adjustment rates and to explore a method suitable for estimating the confidence interval of age-adjusted prevalence in cervical cancer screening studies in Jiangsu Province. Methods The binomial distribution normal approximation method, the Gamma distribution method and the “exact probability method” were used to estimate the interval of age adjustment rate. The statistical simulation was used to investigate the interval coverage and width of the three methods under various rates and positive numbers . Results When the sample size is small (the number of positives is small), the coverage of the exact probability method is better than the Gamma distribution method in terms of theoretical confidence interval and interval width, and the coverage of the two methods is obviously better than the normal approximation ; With the increase of positive number, the deviation of the coverage rate and width of the three methods gradually became smaller and the differences among the methods also gradually reduced; when the positive number increased to above 30, the coverage of the exact probability method and the normal approximation method Of the deviation are within ± 1%, at this time the width of the interval between the two approaches; and Gamma distribution coverage to less than 1% deviation, the total number of positive requirements of 100 or more. No matter whether the sample structure deviates slightly from the overall composition, the above laws are valid. Conclusion Considering the range coverage, the width of the interval and the convenience of calculation, it is suggested that the exact probability method should be used to calculate the confidence interval of the adjustment rate when the total positive number is less than 30. When the total positive number is greater than or equal to 30, the normal approximation method is adopted.