本文采用复合分位数回归(CQR)的估计方法对利率期限结构模型贴现率的估计与预报进行研究。实证结果表明,在小样本情形下复合分位数回归方法要比最小二乘法和分位数回归法稳健性更强,表现出对参数估计的稳定性;对于到期期限不同的债券利用复合分位数回归得到的预报结果更加精确。
In this paper,we use the composite quantile regression(CQR) estimation method to study the estimation and forecast of the term structure of an interest rate model.The empirical results indicate that the CQR method is much more robust than least square regression methods for a small sample.The forecast result of the CQR method is much more accurate than any other methods for a range of bonds with different maturities.