线性回归模型的误差项不服从正态分布或存在多个离群点时，可以将残差秩次的某些函数作为权重引入估计模型来减少离群点的不良影响。
When multiple outliers occur in linear regression model or the distribution of residuals is not normal, we can use residuals rank as weight function to get some resist estimator.
通过逐步回归分析方法剔除次要影响因素，并采用卡尔曼滤波方法动态预测回归残差项。
The sequential regression analysis was adopted to screen off the secondary factors, and the Kalman filtering technique was used to estimate innovation coefficients of the model dynamically.
回归模型的残差项反映了对被解释变量有影响但未列入解释变量的因素所产生的噪音，这部分噪音可由时间序列模型进行拟合。
The residual term in the regression model is the noise generated by the omitted variables that influent dependent variable in the model. The time series model can fit this noise.
应用推荐