【摘要】 目的 应用Hurdle模型分析具有零膨胀数据阳性过敏原数的影响因素,为预防过敏反应及过敏性疾病提供依据。方法 收集2014年就诊变态反应科的患者1 855例,以问卷调查方式获得患者基本信息,测定血清过敏原特异性IgE获得阳性过敏原数;应用logit-NBHurdle模型分析阳性过敏原数的影响因素。结果 在就诊患者中,阳性过敏原数为0的患者占72.29%,数据具零膨胀现象。两次Voung检验的V值分别为3.90、2.77,P值均小于0.05,logit-Poisson Hurdle和logit-NB Hurdle优于传统计数模型;logit-Poisson.Hurdle和logit-NB Hurdle的似然比检验LR=205,P<0.05,结合AIC=1 898.6、AICC=1 900.3、BIC=2 042.4的结果,选择logit-NB Hurdle为最优模型。Logit部分结果为女性比男性更易发生过敏反应;零截尾负二项部分结果为过敏原数目与文化程度、母亲有过敏史有关,文化程度低、母亲有过敏史的患者过敏原数多。结论 分析零膨胀过离散数据前需要进行最优模型的选择;女性是过敏反应的好发人群,预防人群的阳性过敏原数多更应关注母亲有过敏史及文化程度低的人群。更多还原

【Abstract】 Objective To determine the influencing factors of positive allergen number with zero-inflated data using Hurdle model in order to provide scientific evidence for preventing allergy and allergic disease.Methods The basic information of 1 855 patients who visited allergic apartment in 2014 was collected,the number of specific IgE positive allergen from serum allergen was tested,and logit-NB Hurdle model was used to investigate the influencing factors of positive allergens number.Results The positive allergens number of 72.29%patients was 0.Vuong test(V= 3.90 and 2.77,P<0.05) suggested that Hurdle model was better than traditional counting model,according to LR test and AIC,BIC,and AICC values,and logit-NB Hurdle model was chosen.The result of logit part was that women were more likely to produce allergic reaction than men.The result of zero truncated negative binomial part was that mothers having high education degree were likely to have less positive allergens than those having low education degree,and mothers having a history of allergy could increase the number of positive allergen.Conclusion Analysis of zero expansion of discrete data needs to select the optimal model at first.Women are more likely to produce allergic reaction.People with low education degree and whose mother has a history of allergy should be paid more attention to prevent more allergen numbers.更多还原