Hello,
I'm looking to estimate a Poisson fixed effects model with an endogenous variable. I think the correct approach using GMM is specified in Example 12 of the STATA Manual entry for GMM (Page 32 in http://www.stata.com/manuals13/rgmm.pdf). Here is the entry and the code in the example:
"Here we refit the model, treating x3 as endogenous and x1 and x2 as strictly exogenous. Our moment-evaluator program is
program gmm_poiend
version 13
syntax varlist if, at(name)
quietly {
tempvar mu
matrix score double ‘mu’ = ‘at’ ‘if’, eq(#1)
replace ‘mu’ = exp(‘mu’)
local mylhs : coleq ‘at’
local mylhs : word 1 of ‘mylhs’
replace ‘varlist’ = ‘mylhs’/‘mu’ - L.‘mylhs’/L.‘mu’ ‘if’
}
end
Now we call gmm using x1, x2, and L2.x3 as instruments:"
The example invokes the program with the following commands:
use http://www.stata-press.com/data/r13/poisson2
gmm gmm_poiend, nequations(1) vce(cluster id) onestep parameters(y:x1 y:x2 y:x3) instruments(x1 x2 L2.x3, noconstant)
However, when I follow these steps to reproduce the example, I get an error message. My output is below:
. use http://www.stata-press.com/data/r13/poisson2
. program gmm_poiend
1.
. version 13
2.
. syntax varlist if, at(name)
3.
. quietly {
4.
. tempvar mu
5.
. matrix score double ‘mu’ = ‘at’ ‘if’, eq(#1)
6.
. replace ‘mu’ = exp(‘mu’)
7.
. local mylhs : coleq ‘at’
8.
. local mylhs : word 1 of ‘mylhs’
9.
. replace ‘varlist’ = ‘mylhs’/‘mu’ - L.‘mylhs’/L.‘mu’ ‘if’
10.
. }
11.
. end
. gmm gmm_poiend, nequations(1) vce(cluster id) onestep parameters(y:x1 y:x2 y:x3) instruments(x1 x2 L2.x3, noconstant)
error calling gmm_poiend at initial values
‘mu’ invalid name
r(198);
I'm new to GMM and so please advice what I should do to make the example run smooth. Are there alternative approaches to estimate a fixed effects count model with an endogenous variable?
Best wishes,
Venkat
I'm looking to estimate a Poisson fixed effects model with an endogenous variable. I think the correct approach using GMM is specified in Example 12 of the STATA Manual entry for GMM (Page 32 in http://www.stata.com/manuals13/rgmm.pdf). Here is the entry and the code in the example:
"Here we refit the model, treating x3 as endogenous and x1 and x2 as strictly exogenous. Our moment-evaluator program is
program gmm_poiend
version 13
syntax varlist if, at(name)
quietly {
tempvar mu
matrix score double ‘mu’ = ‘at’ ‘if’, eq(#1)
replace ‘mu’ = exp(‘mu’)
local mylhs : coleq ‘at’
local mylhs : word 1 of ‘mylhs’
replace ‘varlist’ = ‘mylhs’/‘mu’ - L.‘mylhs’/L.‘mu’ ‘if’
}
end
Now we call gmm using x1, x2, and L2.x3 as instruments:"
The example invokes the program with the following commands:
use http://www.stata-press.com/data/r13/poisson2
gmm gmm_poiend, nequations(1) vce(cluster id) onestep parameters(y:x1 y:x2 y:x3) instruments(x1 x2 L2.x3, noconstant)
However, when I follow these steps to reproduce the example, I get an error message. My output is below:
. use http://www.stata-press.com/data/r13/poisson2
. program gmm_poiend
1.
. version 13
2.
. syntax varlist if, at(name)
3.
. quietly {
4.
. tempvar mu
5.
. matrix score double ‘mu’ = ‘at’ ‘if’, eq(#1)
6.
. replace ‘mu’ = exp(‘mu’)
7.
. local mylhs : coleq ‘at’
8.
. local mylhs : word 1 of ‘mylhs’
9.
. replace ‘varlist’ = ‘mylhs’/‘mu’ - L.‘mylhs’/L.‘mu’ ‘if’
10.
. }
11.
. end
. gmm gmm_poiend, nequations(1) vce(cluster id) onestep parameters(y:x1 y:x2 y:x3) instruments(x1 x2 L2.x3, noconstant)
error calling gmm_poiend at initial values
‘mu’ invalid name
r(198);
I'm new to GMM and so please advice what I should do to make the example run smooth. Are there alternative approaches to estimate a fixed effects count model with an endogenous variable?
Best wishes,
Venkat
Comment