 Notice: On April 23, 2014, Statalist moved from an email list to a forum, based at statalist.org.

# st: -gmm- substitutable expression with an AR process

 From Daniel Borowczyk Martins To statalist@hsphsun2.harvard.edu Subject st: -gmm- substitutable expression with an AR process Date Wed, 10 Aug 2011 15:55:13 +0100

```Hello stata users,

I have a question regarding the use of -gmm-'s substitutable expression option.

Consider a 3 year time series data set, with monthly observations and
variables "y x year"

I want to run the following model " y = xb  year2 year3" where year#
are year dummies.

I assume the residual is AR(1) with autoregressive parameter rho and
therefore take quasi-differences to the model.

x is assumed endogenous and I therefore use its own and y's second lag
as instruments.

I use stata's -gmm- to implement my estimation procedure.

tab year, gen(year)

local yd "year2 year3"

gmm (y - l.y - (1 - {rho})*{b0} - (x1*{b1} - l.x1*{rho}*{b1}) - ({xb:
`yd'} - {rho}*{xb: l.(`yd')})), instruments(`yd' l(1/2).(y x))

I get the following error message "xb already declared a linear combination".

I can get -gmm- to do what I want but it requires writing the model
expression without the xb option, which is a bit of a pain, esp. with
a higher order AR process.

gmm (y - l.y - (1-{rho})*{b0} - (x1*{b1} - l.x1*{rho}*{b1}) -
({tau2}*year2 + {tau3}*year3 - {rho}*{tau2}*l.year2 -
{rho}*{tau3}*l.year3)), instruments(`yd' l(1/2).(y x))

Is there anyway to use xb, or the moment evaluator expressions to do
what I want in a parse fashion?