Using -suest- with first-differences and factor variables (?)

[Alex Alain]

Hi.

I am trying to find the standard error of a product of two coefficients that come from different regressions.

Therefore I am trying to use a combination of the -suest- and -nlcom- commands.

However, I am running into an error when I use the -suest- command. I keep getting the error: "estimation sample of the model saved under m1 could not be restored r(198)".

I have used -estimates save- and -estimates use- and -estimates sto- to make sure the two estimate stores are loaded.
-est dir- shows me that both my models are loaded.

I don't understand why I am getting the error. Could it be that -suest- is not compatible with D. commands (for first-differencing) or perhaps factor variables?

Really have spend so many hours online to try to find an answer.

As an alternative, is there an easy way to calculate the standard error of the product of 2 coefficients manually?

Many thanks!

[Tiago Pereira]

It would be easier to help with a detailed example of your codes. If you need 95% CIs, have you tried bootstrapping?

[Alex Alain]

Hi Tiago,
Thanks for your response.

reg D.(ln_agr_prod spei_05) i.year i.country_id
est save m1

reg D.(ln_GDPcap_WDI agcomp_lag spei_05) i.year i.country_id
est save m2

est use m1
est sto m1

est use m2
est sto m2

est dir 

suest m1 m2, cluster(country_id)

-est dir- confirms both stores are loaded.

I then get this r(198) error so don't even have the chance to run -nlcom-

Any thoughts?