一种基于分位数回归的结构突变检验

A Verification of Structural Break Based on Quantile Regression

很多模型结构突变研究都是在条件均值下开展的,分析条件分位数下的结构突变同样重要。文章针对线性分位数回归模型中,参数结构为临时性改变,提出了一种移动估计检验方法,并给出了在原假设条件下的极限分布,模拟结果表明该统计量具有良好的检验功效,同时检验功效与断点起始位置、终点位置、残差的分布、断点程度、模型形式、条件分位数、样本量等因素都存在紧密联系。

Many studies on model structure break are carried out under the conditional mean, but to analyze structural break under the conditional quantile is equally important. This paper aims at the temporary change of structural parameters in the linear quantile regression model to propose a mobile estimation method, and also gives limit distribution under the original hypothesis. The simulation results show that the statistic has an excellent test effect, and that at the same time, the test efficacy is closely relat- ed to such factors as breakpoint start position, end position, residuals distribution, breakpoint deviation, model form, conditional quantile, sample size and so on.