回归诊断在城市空气质量预报中的应用研究

An Application of Regression Diagnosis to City Air Quality Forecasting

城市空气质量回归预报模型的残差分布存在着不对称现象 ,它是由高杠杆点引起。这些高杠杆试验点的残差存在着统计天气预报意义上的不合理性 ,导致回归系数Ls估计的误差 ,从而引起预报的误差。针对这些问题提出了城市空气质量的回归诊断预报模型。实例计算说明 ,回归诊断预报模型要优于常规回归预报模型。进一步分析指出 ,城市空气质量回归预报模型的不合理性并非个别例子的特殊性所造成 ,而是由模型的数学特点所决定 ,因此城市空气质量的回归诊断预报模型具有普遍意义。

There exists asymmetry characteristic of residual distribution in the regression model of city air quality forecast. It is caused by some high leverage cases. The residues of these high leverage cases have no rationality in the sense of statistical weather forecasting, and the errors of Least Square Estimation (LS) of the regression coefficient are occurred. So the errors of city air quality forecast are occurred. Thus the regression diagnosis prediction model for city air quality forecasting is proposed. The mathematical proving and the calculation of the examples show that this new model is superior to the general regression prediction model. It is illustrated by means of further analysis that rationality of regression model of city air quality is not caused by some examples, but by mathematical characteristics of the model.