含区间数据Gamma分布变量的因子分析

目的　建立含区间数据Gamma分布变量的因子分析方法,并应用于SARS潜伏期的影响因素分析。方法　采用EM算法求解基于含区间数据Gamma分布的广义线性模型(GLM)参数的极大似然估计(MLE)。采用所建立的GLM方法对含区间数据SARS潜伏期进行单因子、多因子分析。结果　采用EM算法,得到了含区间数据Gamma分布GLM中各个参数的MLE、近似的标准误、以及参数和模型的显著性检验结果。用于SARS潜伏期影响因素分析,发现不同接触方式、不同疫区对SARS潜伏期有明显影响;随着年龄增长,SARS潜伏期均值呈上升趋势;医务人员的潜伏期明显短于非医务人员。结论　基于EM算法的MLE方法对于含区间数据Gamma分布GLM参数的估计是强健的,GLM方法可以用于含区间数据SARS潜伏期的影响因素分析。

Factorial Analysis of Dependent Variable with Interval Data Fitted to Gamma Distribution

Purpose To develop a method to make facto ri al analysis of dependent variable with interval data fitted to Gamma distrib ut ion and conduct it to explore factors that influencing the incubation period of severe acute respiratory syndrome (SARS). Methods EM algorithm was employed to solve the maximum likelihood estimation (MLE) of parameters of general linear model (GLM) based on Gamma distribution with interval data,then univariate and multivariate anal yses were applied to explore factors that influen-cing SARS incubation period. Results We obtained MLE,the approximate standard error of each parameter of GLM,and the results of significant test of GLM.The results from factorial analysis of SARS incubation period showed that both contact patte rn and epidemic area had significant impacts on the incubation period;older pati ents had a longer incubation period;and the incubation period of the medical per sonnel was obviously shorter than that of the non-medical personnel. Conclusions MLE based on EM algorithm is robust for par ameter estimation of GLM based on Gamma distribution with interval data.GLM meth od can be used to explore factors that influencing SARS incubation period with i nterval data.