摘要
目前,虽然已经有好几种统计方法可用于制定多元医学参考范围,如多元正态容许区间法、多元回归法和多维标度法等,但由于这些方法计算复杂,结果表达不够直观,或对资料有较严的要求(如多元正态分布),影响了多元医学参考范围的普及和应用。为了克服上述方法的缺点,本文提出了制定多元医学参考范围的非参数方法。该方法是制定一元医学参考范围的百分位数法在多元情况下的推广,具有非参数统计方法的种种优点,如对资料的分布没有任何要求,道理浅显,计算简便,结果表达直观等,是医学研究工作者制定多元医学参考范围较理想的统计方法。
Some statistical methods have been used to esti- mate multivariate medical reference regions,e.g. multivariate Normal tolerance region,multidimen- sional scaling,etc.Because of the defects of these me- thods,i.e.complexity of calculation,indirect result- expression,and requirement of some restriction of the data,e.g.multivariate Normality,an effect has been produced on the application of multivariate medical reference regions to medical practice.To overcome the above problemes,this paper present a nonparametric statistical method to estimate mult- ivariate medical reference regions,which is the ge- neralization of percentile method used in estimating univariate reference region Compared with other methods,the advantages of the nonparametric statis- tical method are distribution-free,easy to learn and understand,simplicity of calculation,and direct result- expression.It is a relatively ideal statistical method for doctors to estimate multivariate medicalreference regions in practice.
