Standard error for product of two coefficients from different regressions

[Alex Alain]

Hello!

I am looking for a way to get the standard error of the result when I multiply two regression coefficients from different regressions.

So, I run regression_1, lets say:
y1=ax1+bx2+u

Then, I run regression _2, lets say:
y2=cx3+dx4+v

Now, for the purpose of my study, I want to multiply a by c, but I would also like to get a standard error for this value.

I have looked around and have seen things about -nlcom- and also -suest- commands but am really not sure of how to use them.
I hope I can get a solution to this (simple?) problem.

Many thanks!

[FernandoRios]

You are on the right track, you need a combination of suest and nlcom

use http://fmwww.bc.edu/RePEc/bocode/o/oaxaca.dta
reg lnwage age female
est sto m1
reg educ age female
est sto m2
suest m1 m2
nlcom [m1_mean]age*[m2_mean]age
nlcom [m1_mean]female*[m2_mean]female

[Alex Alain]

Hi! Thanks for your answer.

However, I forgot to mention I am using FD notation of the sort :

reg D.(ln_agr_prod spei_05) i.year i.country_id, vce(cluster country_id)

Do you think your suggestion will still work? I will try tonight.

[FernandoRios]

you will have to use suest m1 m2, cluster(country id)
But not sure about the other options.

You could also just use GMM adding both models simultaneously

[Alex Alain]

thanks!
However, it seems I am running into another problem.

I forgot to mention, I am running regressions using different datasets. So I thought I could use different frames and pull the estimate stores from the frame. But it seems one cannot do that.

To summarise again:

So, I run regression_1, lets say:
Using dataset1
y1=ax1+bx2+u
est sto m1

then I start a new frame
Using dataset 2
Then, I run regression _2, lets say:
y2=cx3+dx4+v
est sto m2

But -suest- will not let me use m1 from the first frame.

... any ideas?

[FernandoRios]

append both datasets together.

[Alex Alain]

Merging my datasets will be complicated because I have variables with same names and also different IDs.
Is there any other way of saving estimate stores that don't get deleted when doing the -clear- when using the new dataset.

[Alex Alain]

UPDATE:
I managed to append my datasets so that both regressions are using the same data.

However, I am still getting an error= "estimation sample of the model saved under top_regression could not be restored r(198)"

Could it be that -suest- does not work with D. (first difference) command or factor variables?

Getting standard errors for combinations of estimates seems like a pretty simple/common thing to ask for. I don't know why this isn't working.