Standard Error of the Spread Parameter Using GLM

[Danny Gold]

I'm estimating a linear model with a gaussian error that's mean zero with an unknown variance. I am estimating the parameters, including the spread parameter, via maximum likelihood. The model has many covariates, but I'm testing it with a simpler model:
\[y_{it} = \rho y_{it-1} + \sigma\varepsilon_{it}\]

When I estimate with the mlexp command, I'm able to obtain an estimate of the spread parameter and its standard error. When I estimate with

glm y_it y_itm1, f(gaussian)

I get the same \rho coefficient and standard error, and the scale parameter is equal to the \sigma coefficient from mlexp, but I can't find the standard error for the scale parameter estimate anywhere in the summary or postestimation macros. Google and the glm manual have also been no help.

How do you obtain standard errors for the scale parameter estimate from the glm command?

[FernandoRios]

I dont think you can.
For cses like that, you probably want to use a different command, like hetregress. or as you did mlexp. They may be superior to GLM if you want to say something about the Scale parameter.
F

[Danny Gold]

Thanks, Fernando! I wasn't aware of hetregress, and it does exactly what I want it to.