st: IV: calculating manually unbiased standard errors for 2sls.

[Alvaro Angeriz <aa314@cam.ac.uk>]

Dear colleagues,
I'm trying to understand how ivreg calculates standard errors for the 
estimators calculated by 2sls. I haven't been successful so far.

I do as follows

*step 1
reg x z1 z2
predict xhat

* step 2
reg y xhat
scalar alphahat = _coef[_cons]
scalar betahat = _coef[xhat]


scalar n = e(N)
su x
scalar var_x = r(sd)^2
g res_iv_2 = (y - (alphahat + betahat * x))^2
su res_iv_2
scalar sigmau_hat_iv_2 =  _result(18)/(n-2)
scalar sigma_beta_iv = sqrt(sigmau_hat_iv_2/(n*var_x))
di "sigma beta iv = " sigma_beta_iv


Could anyone help me explaining what i'm doing wrong?

Thanks!

Alvaro

*
*   For searches and help try:
*   http://www.stata.com/support/faqs/res/findit.html
*   http://www.stata.com/support/statalist/faq
*   http://www.ats.ucla.edu/stat/stata/