| 现代变量选择方法在青少年近视研究中的应用 |
| |
| 作者姓名： | 海豹 李仕明 刘洛如 申立勇 张三国 李偲圆 李翯 康梦田 孙芸芸 孟博 张庆昭 |
| |
| 作者单位： | 1. 中国科学院大学数学科学学院, 北京 101408;
2. 首都医科大学附属北京同仁医院眼科中心, 北京 100730;
3. 河南安阳市眼科医院, 河南 安阳 455000 |
| |
| 基金项目： | 国家973重点基础研究发展计划项目(2011CB504601)资助 |
| |
| 摘 要： | 通过分析一组医学数据挖掘出影响青少年近视的关键因素,建立青少年近视患病概率预测模型.数据集主要由两部分组成:一是青少年眼睛的医学测量数据,二是由生活学习习惯调查问卷得到的数据.采用几种现代统计学方法,并利用ROC曲线得到较优的患病概率模型.结果表明,性别、眼轴长度、角膜曲率、工作日睡眠时间、不戴眼镜远视力、远距离调节反应等因素对青少年近视有重要的影响作用,并由此建立预测模型.
|
| 关 键 词： | 变量选择 logistic回归 Lasso MCP ROC曲线 |
| 收稿时间： | 2014-09-25 |
| 修稿时间： | 2015-03-16 |
Juvenile myopia study using modern variable selection methods |
| |
| Authors: | HAI Bao LI Shiming LIU Luoru SHEN Liyong ZHANG Sanguo LI Siyuan LI He KANG Mengtian SUN Yunyun MENG Bo ZHANG Qingzhao |
| |
| Institution: | 1. School of Mathematical Sciences, University of Chinese Academy of Sciences, Beijing 101408, China;
2. Beijing Tongren Eye Center, Beijing Tongren Hospital, Capital Medical University, Beijing 100730, China;
3. Anyang Eye Hospital, Anyang 455000, Henan, China |
| |
| Abstract: | In this work we used some variable-selection techniques to find out the relevant factors that cause adolescent myopia, and established probabilistic models for myopia prediction. The research is based on a medical dataset consisting of two parts: medical measurement data of the youths and data on daily living habits obtained by questionnaire survey. We used some modern variable selection methods and the ROC curve to evaluate different modes. The results show that gender, axial length, corneal curvature, weekday sleeping time, distance vision without glasses, and remote adjustment reaction have important influences on adolescent myopia. |
| |
| Keywords: | variable selection classical logistic regression Lasso MCP ROC curve |
| 本文献已被 CNKI 等数据库收录！ |
| 点击此处可从《》浏览原始摘要信息 |
| 点击此处可从《》下载免费的PDF全文 |
|
