1: 几个变量之间线性化,存在着异方差现像,,,对这些变量分别取自然对数,可以消除异方差现像。
this is so called Box-Cox transformation, it helps stabilize the variance
2: 归系数的解释也会发生变化,这时表示弹性的概念。 , Yes
3: taking logrithm helps make the scales of all the variables comparable, such that when one inverses the matrix, less numerical problem will emerge.
4: 非线性方程取对数的例子我明白! non-linear --> linear relationship