百分位数回归及应用研究

Quantile Regression and Application Research

在进行定量分析时,最小二乘法已经成为一种可信赖的工具。但是运用最小二乘法的条件比较高,在实际问题中,完全满足条件的情况并不多见,那么在应用时就难以得到无偏的、有效的参数估计量。针对上述问题,以OILPLUS公司取暖用燃油消耗的分布为主要研究对象,在进行参数估计时,应用百分位数回归方法,既可以看到采用百分位数回归方法与采用最小二乘法得到的模型显著不同,又可以得到比最小二乘法更为丰富的信息。

In the quantitative analysis, the least square method （OLS） has become a reliable tool. But the conditions for using the least square method are relatively high, in the actual problem, the cases for fully meeting the conditions are rare, then it is difficult to get a unbiased and valid parameter estimator. In this paper, Aiming at these problems,taking the OILPLUS distribution of heating fuel consumption as the main research object, I apply percentile regression method, you can see the quantile regression method and the least square method is significantly different, you can get richer information than the least squares method.