非线性多口径加权算法评估及其应用前景实例分析

Nonlinear calibration weighting algorithm and the prospects of its use:a examplary study

目的验证非线性多口径加权算法是否能够通过最优化技术更加精准地估计总体的实际情况。方法选取山西省常见风湿病学调查中6 424名完整调查资料受访者作为整体,运用口径权数对其总体结构、总体患病人数及样本容量对口径权数的有效性影响进行估计与分析。结果采用1 000个抽样样本分别用普通加权和口径加权方法对总调查人群中的患病人数进行估计,结果发现普通加权算法对总体患病人数估计的误差百分比为5.69%,而口径加权对总体患病人数的误差百分比仅4.71%,结果验证了口径加权对总体患病人数估计的精确度显著优于普通加权。且多口径加权算法对总体分组和总体患病人数的估计值误差均小于普通加权,不同样本容量的患病人数估计值与实际值更加接近。结论多口径加权算法相比普通加权能够更加精准地估计出总体的实际情况,差异小且有一定的稳定性,是一种可用于补充普通加权方法不足的推算总体的方法。

Objective To verify whether nonlinear clibration weighting algorithm can be more accurate in estimating the study population via optimization techniques.Methods A total of 6424 subjects with complete data in Shanxi Provincial Survey on Common Rheumatic Disorders were recruited as the study population.Calibration weighting was used to estimate and analyze the effects of the population structure,the total number of patients and sample size on the validity of calibration weighting.Results One thunsand sampling samples were used to estimate the number of patients in the total survey population using conventional weighting and calibration weighting algorithms,respectively.As a result,the percentage error for total patient number in the study population as estimated by conventional weighting algorithm was 5.69%vs only 4.71%by calibration weighting algorithm.These results demonstrated that calibration weighting algorithm yielded significantly higher accuracy in estimating the total patient number,compared with conventional weighting algorithm.Moreover,the errors of calibration weighting algorithm for overall grouping and estimated patient number in the study population were smaller than those of conventional weighting algorithm,yielding more close-to-actual estimates of patient number in context of different sample sizes.Conclusion Calibration weighting algorithm offers more accurate estimatesof actual situation in the study population than conventional weighting algorithm,with modest differences and favorable stability.Therefore,this algorithm can be used for estimation of study population to complement the defects of conventional weighting algorithm.