[Date Prev][Date Next][Thread Prev][Thread Next][Date index][Thread index]
st: Nonlinear least squares and robust standard errors
Hi Stata Listers,
I am using Stata Version 7 and the -nl- command to do nonlinear least squares. The equation I am estimating looks something like this " Y = T * X1 + (T * D) * X2 + (T * D^2) * X3 + (T * D^3) * X4 + a ton of fixed effects ". I am having some heteroskedasticity problems and I would like to get some robust standard errors for these coefficients.
After running the -nl- program I tried to use -_robust-. I had the following interaction with Stata:
. matrix D=e(V)
. predict double e,residual
. _robust e, v(D) minus(271)
variable T not found
If I understand this correctly (and certainly this is a daring assumption), Stata doesn't like this because my coefficient names are not the same as their corresponding variables. In my specification there is no one-to-one correspondence between independent variables and coefficients as there might me in regular OLS. Is using Sandwich standard errors appropriate in such a specification? If so is there some way to prod Stata to calculate robust standard errors here.
I am also considering using bootstrapped standard errors here, but it took 4 days just to run 50 reps. Do anyone know if using Stata 8 might speed up this process significantly (I have about 4,000 observations)?
Thanks for your help.
Stanford Graduate School of Business
Stanford CA 94305
* For searches and help try: