跨栏模型在确定鼠体蚤丰盛度预测因子研究中的应用  

Application of Hurdle Model in Identifying Predictors for Flea Abundance on Rats

作　　者：尹家祥[1] 董兴齐[1] 

机构地区：[1]云南省地方病防治所,云南大理671000

出　　处：《地方病通报》2010年第5期1-4,共4页Endemic Diseases Bulletin

基　　金：国家自然科学基金项目(81060229);云南省应用基础研究项目(2009CD126);云南省高层次科技人才培引工程(2009CI010)

摘　　要：目的介绍应用跨栏回归模型拟合现场计数资料的一个实例。方法在R统计软件下,使用负二项分布模型和跨栏负二项分布模型分别对横断面调查的计数资料进行拟合,然后比较两种模型拟合的结果。结果跨栏负二项分布模型可以确定影响鼠体染蚤率和染蚤密度的因素,而负二项分布模型仅能确定影响鼠体寄生蚤的因素,但不能区分是影响染蚤率还是蚤密度;另外,跨栏负二项分布模型比负二项分布模型能捕获更多的影响鼠体寄生蚤的因素。结论跨栏负二项分布模型拟合离散型资料比负二项分布模型拟合有优势,而且便于专业上解释。Objective To introduce an example of applying hurdle regression model to fit field count data.Methods A count data from a cross-sectional study was fitted using negative binomial model and hurdle negative binomial regression model in R software,and then the results from two models were compared.Results Hurdle negative binomial model was capable of identifying the predictors for flea prevalence and flea intensity on rat,while negative binomial model was just capable of identifying the predictors for flea abundance on rat and was indistinguishable predictors affecting flea prevalence or flea intensity on rat.In addition,hurdle negative binomial model captured more affecting factors for flea on rat than negative binomial model did.Conclusions Hurdle negative binomial model is preferred to fit discrete data and convenient for interpreting the results.

关 键 词：计数资料 跨栏模型 负二项分布模型 预测因子 鼠 蚤 

分 类 号：R384.3[医药卫生—医学寄生虫学] R181.22[医药卫生—基础医学]