1。首先reg  y x1 x2 [aw=x3]

那个是中括号，

2、加权的变量x3必须事先有，

如果没有，你必须自己得到

3、你可以参考stata7的manual里面的例子

谢谢斑竹。

提示1，我是笔误。实际操作是用的中括号。

提示2，还是不明白。我该如何得到这个weight?

提示3，我现在版本是8，找了半天帮助里面也没找到例子。

明天要见导师，现在还没做出来，请帮帮忙啊，

多谢了！！

难道是这个命令

qreg y x1 x2 x3 x4 x5

check stata help, weight is defined in residual variance=sigma^2/w.

So if variance =x3*sigma^2, then weight is 1/x3

generate a new variable called wx=1/x3, and use

reg y x1 x2 x3 [aweight=wx]

这是greene书上的例子

use "d:\data\dta\greena\table51-greene.dta", clear
rename x1 age
rename x2 income
rename x3 exp
rename x4 ownrent
rename x5 selfemp

generate incomesq = income^2
drop if exp==0

*save chapter12hetero,replace

regress exp age ownrent income incomesq
rvpplot income, xlab(0 2 to 12) xline(2 4 6 8 10) ylab(-500 0 to 2000) yline(0 500 1000 1500)
regress exp age ownrent income incomesq, robust

matrix d = vecdiag(e(V))
matrix v =  cholesky(diag(d))
matrix s = sqrt((72-5)/72)*vecdiag(v)
matrix list s

regress exp age ownrent income incomesq, hc2

whitetst
hettest
bpagan income incomesq

wls0 exp age ownrent income incomesq , wvar(income) type(abse) noconst          /* 12.3a */
wls0 exp age ownrent income incomesq , wvar(incomesq) type(abse) noconst        /* 12.3b */
wls0 exp age ownrent income incomesq , wvar(income incomesq) type(e2) noconst   /* 12.3c */
wls0 exp age ownrent income incomesq , wvar(income incomesq) type(abse) noconst /* 12.3d */
wls0 exp age ownrent income incomesq , wvar(income incomesq) type(loge2)        /* 12.3e */
wls0 exp age ownrent income incomesq , wvar(income incomesq) type(xb2)          /* 12.3h */
regress exp age ownrent income incomesq [aw = 1/income]
regress exp age ownrent income incomesq [aw = 1/incomesq]

regress exp age ownrent income incomesq

predict e, resid
generate ee=e^2
regress ee income incomesq, noconst

predict p1
regress exp age ownrent income incomesq [aw = 1/p1]

generate abse=abs(e)
regress abse income incomesq, noconst

predict p2
regress exp age ownrent income incomesq [aw = 1/p2]

generate logee=log(ee)
regress logee income incomesq

predict p3
replace p3 = exp(p3)
regress exp age ownrent income incomesq [aw = 1/p3]

generate loginc = log(income)
regress logee loginc

predict p4
replace p4 = exp(p4)
regress exp age ownrent income incomesq [aw=1/p4]

reghv exp age ownrent income incomesq, var(loginc)

regress exp age ownrent income incomesq

predict p
generate p5 = p^2
regress exp age ownrent income incomesq [aw=1/p5]
reghv exp age ownrent income incomesq, var(income incomesq)

谢谢楼上各位，我做出来了！多谢：）

楼主是怎么解决的啊？ 最近碰到同样的问题了。

你好，请问stata中权重变量在众多自变量里该选择哪一个呢？谢谢你。

先做wald test，得到回归的拟合值 那个拟合值就weight