基于贝叶斯复合分位数回归的参数估计及应用

针对传统最小二乘估计易受异常点干扰及稳健性较差的问题,建立了基于复合分位数回归估计的数据拟合预测模型。为了克服复合分位数回归在估计参数时忽视了参数的不确定性,致使估算出的参数精度不够高的缺点,将贝叶斯分析法与复合分位数回归相结合,提高了参数的估算精度。实证分析表明贝叶斯复合分位数回归估计优于复合分位数回归估计,而复合分位数回归估计优于传统最小二乘估计,值得工程技术人员借鉴。

Parameter estimation and application based on Bayesian composite quantile regression

Aiming at the problem that the traditional least squares estimation is vulnerable to outliers and its robustness is poor, the data fitting and forecasting model based on the composite quantile regres-sion estimation is established. In order to overcome the composite quantile regression shortcomings in the estimation of parameters ignore parameter uncertainty, resulting in the disadvantages of estimated parame-ters precision is not very high. By combining the Bayesian analysis method and the composite quantile re-gression, the estimation accuracy of the parameters is improved. The empirical analysis shows that the Bayesian composite quantile regression estimation is better than the composite quantile regression estima-tion, and the composite quantile regression estimation is better than the traditional least squares estima-tion, and it is worth learning from the engineering and technical personnel.