adaptive LASSO logistic回归模型应用于老年人养老意愿影响因素研究的探讨  

Influencing Factors on Elder's Preference for Supporting: Application of an adaptive LASSO logistic Model

作　　者：韩耀风[1,2] 覃文峰 陈炜[1,2] 李博涵[1,2] 滕伯刚[1,2] 方亚[1,2] 

机构地区：[1]厦门大学公共卫生学院,361102 [2]卫生技术评估福建省高校重点实验室

出　　处：《中国卫生统计》2017年第1期18-22,共5页Chinese Journal of Health Statistics

基　　金：厦门大学大学生创新创业训练计划项目(2015Y0827)

摘　　要：目的探讨adaptive LASSO logistic回归模型在老年人养老意愿影响因素研究中的应用。方法基于厦门市60岁及以上老年人口的多阶段整群抽样调查数据,建立老年人养老意愿影响因素的adaptive LASSO logistic回归模型,通过交叉验证法选择模型中的调和参数λ;通过与全变量和逐步logistic回归结果的比较,探讨adaptive LASSO logistic回归模型的优势。结果共纳入1244名老年人,其养老意愿为家庭养老、社区居家养老和机构养老的比例分别为70.0%、21.1%和8.9%。交叉验证法选择的λ为0.018;此时adaptive LASSO logistic回归模型纳入的自变量为居住地、年龄、婚姻状况、文化程度、子女数、每月退休金收入、公费医疗和住院情况;BIC和AIC分别为1931、1888,均低于全变量logistic回归(2077、1923)和逐步logistic回归(2025、1912)。结论 adaptive LASSO logistic回归模型可用于老年人养老意愿影响因素研究。老年人的养老意愿受多个因素影响。Objective This study was aimed to analyze influencing factors on elder＇s preference for supporting in case to provide evidence for diversified elderly supporting. Methods This article applied an adaptive LASSO logistic model to the multi-stage cluster sampling data of the population aged 60 or older in Xiamen to analyze influencing factors on elder＇s preference for supporting. Cross validation methodwas used to choose λ for adaptive LASSO logistic model. In addition,w e evaluated the model fitting of adaptive LASSO logistic model by comparing the BIC and AICwith full logistic model and stepw ise logistic model. Results The cross validation method resulted in λ = 0. 018 for adaptive LASSO logistic model,inwhich variables retainedwere residence,age,marital status,education level,number of children,the monthly pension income,medical insurance and hospitalization. BIC of adaptive LASSO logistic model,full logistic model and stepw ise logistic modelwere 1931,2077 and 2025respectively. And AIC of the three modelswere 1888,1923 and 1912 respectively. Conclusion Compared to full logistic model and stepw ise logistic model,adaptive LASSO logistic modelwas the best fitting model for influencing factors on elder＇s preference for supporting. Adaptive LASSO logistic model could be used to analyze influencing factors on elder＇s preference for supporting. Therewere multiple factorswhich influenced elder＇s preference for supporting.

关 键 词：ADAPTIVE LASSO LOGISTIC回归模型 养老模式 影响因素 

分 类 号：R195[医药卫生—卫生统计学] D669.6[医药卫生—卫生事业管理]