带多个变点的逐段连续线性分位数回归模型及应用

发布时间：2018-06-17 04:07

本文选题：多个变点 + 逐段连续线性　；参考：《数量经济技术经济研究》2017年08期

【摘要】：研究目标:建立具有多个变点的逐段连续线性分位数回归模型(Continuous Piecewise Linear Quantile Regression with Multiple Change Points,CPLQR)。研究方法:先通过LASSO和广义贝叶斯信息准则确定变点个数,再通过线性化技巧来估计变点的位置与回归系数。研究发现:新方法能够同时确定变点个数、估计变点位置和回归系数,而且具有较强的稳健性;应用该方法于年龄和身体质量指数之间关系,进一步证实了模型的实用性。研究创新:新方法能够处理多个变点的问题,通过LASSO和广义贝叶斯信息准则确定变点数目,避免了主观判断的弊端;借助线性化技巧,解决了目标函数在变点处不可导问题。研究价值:本文结果将为分析经济、金融、医药和生物等学科中存在结构变化的数据提供强有力的研究工具。
[Abstract]:Objective: to establish a piecewise continuous linear quantile regression model with multiple variable points, continuous Piecewise Linear quantitative with multiple change points (CPLQR). Methods: the number of change points is determined by LASSO and generalized Bayesian information criterion, and then the location and regression coefficients of change points are estimated by linearization technique. It is found that the new method can determine the number of change points at the same time, estimate the location of change points and the regression coefficient, and has strong robustness, and further proves the practicability of the model by applying this method to the relationship between age and body mass index. Research innovation: the new method can deal with the problem of multiple change points, determine the number of change points by LASSO and generalized Bayesian information criterion, avoid the disadvantages of subjective judgment, and solve the problem of non-derivation of objective function at the change point by means of linearization technique. Research value: the results of this paper will provide a powerful research tool for the analysis of structural changes in economics, finance, medicine and biology.
【作者单位】：湖南大学金融与统计学院;
【基金】：国家自然科学青年基金项目(11401194) 湖南省自然科学青年基金项目(2017JJ3021) 湖南大学中央高校基本科研业务费专项资金(227201305039)的资助
【分类号】：F224

【参考文献】