quantile regression model coefficient CIs using the empirical variance distribution approximation

This paper investigates the application of the empirical variance distribution (evd) function to estimate the confidence interval bounds of the quantile regression model coefficient estimates for homoscedastic unweighted data.<br>It can be seen that for this modest sample size, the bootstrap estimator and the two evd based confidence interval estimators exhibit nominal 95% coverage for homoscedastic iid cases in slope estimates and &gt;93% in the intercept estimates. For extreme quantiles in smaller samples or when homoscedastic iid error is not present, the coverage performance is lower.本文针对同方差未加权数据，探究了经验方差分布（Empirical Variance Distribution, EVD）函数在估计分位数回归模型系数估计值的置信区间边界中的应用。 结果表明，在该适中样本量下，针对同方差独立同分布场景，自助法估计量与两种基于EVD的置信区间估计器在斜率估计值上达到了名义95%的覆盖率，截距估计值的覆盖率则超过93%。而在小样本下的极端分位数场景，或是不存在同方差独立同分布误差项时，其覆盖率表现均有所下降。