Manipulating results stored by commands

[Felix Holub]

Is there a way to change a stored result? For example, after a regression I would to change the value of e(N) itself to then apply some postestimation commands. To provide some context, I would like to use

-mfxrcspline-

from SSC with my own standard errors which I obtain from a combination of bootstraps.

[Clyde Schechter]

You cannot change what is stored in e(). However, you can write code that stores new "estimates" that are more to your liking. See -help ereturn-.

[daniel klein]

You cannot change what is stored in e(). However, you can write code that stores new "estimates" that are more to your liking. See -help ereturn-.

Technically, this statement is incomplete. It is true that e(b), e(V) and e(Cns) can only be altered using an eclass program and ereturn repost. Other results, stored in e() cannot be [directly, i.e., outside an eclass program] changed in Stata, but you can use Mata. Here is an example.

sysuse auto
regress price mpg i.foreign
display e(N)
mata : st_numscalar("e(N)", 42)
display e(N)

In each case, you should really be sure that you know what you are doing. Messing with Stata's returned results might have unexpected consequences and might give you results that you should be able to obtain for good reason.

Best
Daniel

[Felix Holub]

Thank you Daniel and Clyde, you pointed me to the relevant issues. For future reference, I present my solution here, using

-erepost-

available from SSC:

sysuse auto, clear
reg price mpg
matrix list e(V)
matrix  V = (1, 1 \ 1, 1 )
erepost V = V, rename
matrix list e(V)

[Felipe Lozano]

Hi Felix,

I am trying to do something similar to include both estimates and confidence intervals from a regression and from a bootstrap, with the aim of exporting them to latex though esttab o estout. Have you found a way to alter the e-class elements feeding esttab/estout? That would be very handy.
Best,
Felipe L.