暴露评估中处理未检出值的多检出限回归方法及其应用  

Application of multiple limits regression in dealing with nondetects with multiple limits of detect

作　　者：余小金[1] 刘沛[1] 闵捷[1] 张宏[1] 陈炳为[1] 陈启光[1] 魏永越[2] 陈峰[2,1] 

机构地区：[1]东南大学公共卫生学院流行病学与卫生统计学系,南京210009 [2]南京医科大学公共卫生学院

出　　处：《卫生研究》2008年第6期745-747,共3页Journal of Hygiene Research

基　　金：国家“十一五”科技支撑计划重大项目(No.2006BAK02A01);江苏省卫生厅预防医学科研课题资助(No.Y200702)

摘　　要：目的探索暴露评估中多检出限回归在存在多个检出限的情况下未检出值数据填补中的应用。方法介绍多检出限回归方法中次序统计量的计算方法,并将方法应用与于处理2004年全国淡水鱼重金属镉残留量数据,对未检出值进行填补,并将填补后的参数估计与简单替换的方法进行比较分析。结果多检出限回归填补数据后估计的2004年全国镉残留平均值(mg/kg)为0.01509,3种不同替换值(0、1/2和1倍检出限)的估计结果分别为0.014759、0.015270和0.015781。结论多检出限回归填补估计方法利用检出值的信息对未检出值进行填补,较简单替换法更为合理。通过SAS宏程序可以方便得以实现,值得在食品暴露评估中推广应用。Objective To investigate the application of Multiple Limit Regression （MR）, that is alternative to regression on order statistics, in imputing nondetacts below multiple detect limits. Methods The Multiple limits regression method was used to dealing with the nondetects in cadmium residue in Fish food in 2004 from China markets; The nondetects are imputed with the predicted values based on MR; The results of estimates are compared with those from simple substitution methods. The analysis was performed using self compiled SAS macro code. Results The mean estimated by MR methods is 0.01509mg/kg, the means from substitution methods with 0, 1/2LOD and LOD are 0.014759, 0.015270,0.015781 respectively. Conclusion MR is a worthy recommendatory method in handling nondetects with multiple limits of detect for its rationality and can be completed conveniently by compiled SAS macro code.

关 键 词：未检出值 多检出限回归 次序统计量回归 镉 

分 类 号：R155.51[医药卫生—营养与食品卫生学] TS201.6[医药卫生—公共卫生与预防医学]