Wald Test for two different samples

[Syed Raza]

Hello everyone,

I am running WALD test for two different samples which means do different codes. I ran OLS in each sample and i get one coefficient in each equation. I want to test to whether both coefficients are equal or not but problem comes later when i use test command. Here comes the procedure i ran the regression, store the estimates and ran the second regression, i restore the estimates of first regression and use

suest father=father1, coefl

father is coefficient from first regression and father1 is coefficient from second regression. Stata gives me error father1 not found. How can i test the whether these coefficients are same or not. Thanks

[Andrew Musau]

See this example where I am testing the equality of the coefficients on mpg from the two regressions

sysuse auto
qui reg price mpg
est sto mod1
qui reg price mpg weight
est sto mod2
suest mod1 mod2
test _b[mod1_mean:mpg] =  _b[mod2_mean:mpg]

[Syed Raza]

Andrew Musau This does not work. I am running two different do files, means two different samples so it gives me an error at this command suest "estimation sample of the model saved under mod1 could not be restored"

[Weiwen Ng]

Andrew Musau This does not work. I am running two different do files, means two different samples so it gives me an error at this command suest "estimation sample of the model saved under mod1 could not be restored"

That is the problem. I would append both files, e.g.

use file1.dta
append using file2.dta, generate(file)

reg outcome father if file == 0
estimates store model1

reg outcome father1 if file == 1
estimates store model2

suest model1 model2
test _b[model1_mean:father] = _b[model2_mean:father1]

Basically, Stata has to be able to mark the estimation sample in both datasets for suest to work. Stata can only hold one dataset in memory at a time. So, when you referred to your first model, Stata could not find the estimation sample. It needs to see that to run suest.