how to use stored estimates for contrasts in fixed effects regressions?

[kann sriki]

Hello All,
This must be a simple problem - but I have searched for a day and somehow cannot find the answer...
I am trying to do across-sample contrasts in fixed effects regressions - but I cannot find the syntax for it:

xtreg y x1 x2 if subsample==1, fe
est store M1

xtreg y x1 x2 if subsample ==2, fe
est store M2

I want to do something like
test M1_x1 - M2_x2
test M1_x2 - M2_x2

I can manually calculate the difference, but was hoping there is a stata command for me to do this?
Many Thanks.

[Clyde Schechter]

Actually, you can't do this. The command that is usually used for cross-regression estimation, -suest- does not support -xtreg-.

However, you can get what you want in a completely different way:

xtreg y i.subsample##c.(x1 x2), fe

The coefficient of 2.subsample#x1 will be the difference between M1:x1 and M2:x1, and the coefficient of 2.subsample#x2 will be the difference between M1:x2 and M2:x2.

I am assuming that your variable subsample takes on only the two values 1 and 2.

[kann sriki]

Thanks!