论文部分内容阅读
常用的多元线性回归有一个不足之处,即由于因子之间的相关性以及回归系数间的相互关联,使得回归系数的显著性检验变得比较复杂,除计算相关系数外,还需计算偏相关系数。而且每当选入或删除因子后,各系数都得重新计算,因此计算过程相当复杂。通过正交多项式变换就可以使变换后的因子相互之间无相关性,从而简化检验过程。本文利用四川省30个地区的食管癌死亡率资料,用正交多项式回归法建立了地理坐标与死亡率之间的回归方
Commonly used multiple linear regression has a deficiency, that is, due to the correlation between factors and the correlation between regression coefficients, making the significance test of the regression coefficient becomes more complex, in addition to calculating the correlation coefficient, but also need to calculate the partial correlation coefficient. And each time the factor is selected or deleted, each coefficient has to be recalculated, so the calculation process is quite complicated. The orthogonal polynomial transformation can make the transformed factors have no correlation with each other, thus simplifying the inspection process. This paper uses the data of the mortality rate of esophageal cancer in 30 areas of Sichuan Province and establishes the regression between geographic coordinates and mortality using orthogonal polynomial regression.