Prevent computation of standard errors

[Bram Hogendoorn]

Dear Stata users,

I am running some calculations on a large dataset. The computation time is long. Usually, I would take a subsample from my data to speed up computation time, but this is not possible in my case. Furthermore, I am not interested in the standard errors, only in the point estimates.
Hence my question: is it possible to estimate only the point estimates, and not the standard errors? Currently I am using gsem, but a more general solution would be welcome too.

Thank you,

[Phil Bromiley]

You didn't get a quick answer. Your question is not really clear. You give us no idea of the kind of model you're really estimating. What might work for gsem doesn't necessarily work for other routines and vice versa.

If processing time is a big issue, then I'd see if there are other routines that do what you need rather than gsem. While gsem will estimate an incredible variety of models, often a narrower model can take advantage of structure to give faster estimates. Naturally, using the multiprocessor version of Stata and a faster computer with more RAM also could help a great deal. I don't know off hand of procedures that let you skip standard errors, but there may be some. If such an option exists, it will normally be documented. If it were something very simple like regression, you could program it in Mata, but that requires getting comfortable with Mata (and regression is pretty fast).

[Bram Hogendoorn]

Thank you, Phil. I will restate my question. Is it possible to estimate regression models without estimating the standard errors?
I have searched the documentation in various built-in commands, such as -regress- and -gsem-. The answers seems to be "No, unless you program it yourself."