"Mata object __000002 already exists" after ivreg2

[Denis Alves]

Dear Statalist fellows,

Stata/MP 17, macOS

After loading a csv file and generating some interactions and labelling others, I ran this:

ivreg2 depvar t1##(v1 v2 v3) t2##(v1 v2 v3) t3##(v1 v2 v3) t4##(v1 v2 v3) t5##(v1 v2 v3) (endogreg = iv1 iv2) i.year i.sic, first gmm2s rob endog(endogreg) orthog(iv2) redundant(iv1 iv2)

... which works fine. Then:

estimates store A

, with no errors and correspondent creation of an additional item in the variables list, then:

estimates table A

, which gives the following error:

Mata object __000002 already existsr(110);

This is the first and most important issues which I could not solve after a week searching for an answer.

The second is the obviously weird syntax used for the regression. Using this notation:

ivreg2 depvar (t1-t3)##(v1 v2 v3) (endogreg = iv1 iv2)

... returns the following message:

syntax is "(all instrumented variables = instrument variables)"
invalid syntax
r(198)

I appreciate any help with both, but especially the "already exists" error.

Cheers
Denis

[Denis Alves]

An upgrade to Stata 18 fixed the "already exists" issue. However, I would appreciate your kind help with the syntax

ivreg2 (t1-t3)##(v1 v2 v3) (endogreg= iv1 iv2)

Is it really necessary to type in

ivreg2 t1##(v1 v2 v3) t2##(v1 v2 v3) t3##(v1 v2 v3) (endogreg= iv1 iv2)

?

[Andrew Musau]

ivreg2 is from SSC (FAQ Advice #12). The parsing of parentheses can be tricky in IV regression. You can use fvexpand.

fvexpand (t1 t2 t3)##(v1 v2 v3)
ivreg2 depvar `r(varlist)' (endogreg= iv1 iv2)

[Denis Alves]

Dear Andrew Musau, thank you so much. This is awesome.