predict u , residual
gen usq=u^2
gen lnusq = ln(usq)
reg lnusq X1 X2 X3 X4
predict g
gen h = exp(g)
gen y=Y/sqrt(h)
gen x5=1/sqrt(h)
gen x1= X1/sqrt(h)
gen x2=X2/sqrt(h)
gen x3=X3/sqrt(h)
gen x4=X4/sqrt(h)
reg y v x1 x2 x3 x4 , noconstant
这个异方差修正的代码里为什么多了一个变量x5，x5代表什么呢，是修正之后的随机扰动项吗？，为什么回归要用过原点回归，x5是代表截距了吗