多水平模型在艾滋病流行趋势研究中的应用

Application of Multilevel Model in AIDS Epidemic Trend Research

目的对多水平负二项分布模型在艾滋病病流行趋势研究中应用的优势和不足做探索性研究。方法利用重庆市1994-2009年各区县年度新发艾滋病频数数据以及各区县年度常住人口数,结合多水平建模方法分别拟合Poisson模型和负二项分布模型计算参数并比较。结果负二项分布的超方差系数k=5.887 279,较之Poisson分布模型中的超方差系数k=9.008 424有所减小,但仍大于1且有统计学意义,2种模型拟合结果有差异但不明显。结论多水平负二项分布模型在对艾滋病流行趋势的研究中较之传统的Poisson分布模型无明显差异,2种模型解释的侧重点不同。

Objective To discuss the advantages and disadvantages of the multilevel negative binomial model applicated in AIDS epidemic trend research.Methods Country’s annual HIV epidemic data of Chongqing was used in multilevel Poisson regression model and multilevel negative binomial regression model.Then,calculate the parameters of the models and compare.Results Extra variance coefficient of the negative binomial distribution equal 5.887279,compared with the model of Poisson distribution coefficient 9.008424 has been reduced,but still greater than 1 and statistically significant.Conclusion There was no significant difference between multilevel Poisson regression model and multilevel negative binomial regression model.Two models’ explaination have the different emphases.