what is the correct specification of lagged variables using xtpmg ?

[moh saleh]

Hi everyone,
I'm using xtpmg to run an ardl model on panel data using the pooled mean group estimator. dependent variable: y independent variables x1 and x2
the code line i'm using is similar to a one published in paper which uses ardl(1,1,1)
xtpmg D.y D.x1 D.x2, ec(ec) lr(L.y x1 x2) replace pmg

I'm confused on what would be the appropriate xtpmg specification for different lag orders. using the example above if i want to try ardl(2,0,1) what would be the appropriate code line:
xtpmg D2.y x1 D.x2, ec(ec) lr(L.y x1 x2) replace pmg
Or
xtpmg D(1/2) x1 D.x2, ec(ec) lr(L.y x1 x2) replace pmg
Or
something completely different?

Also, does the lag structure affect the long run lr () specification at all?

If anybody can help or refer me to a paper or guidelines that touches on this, i'd really appreciate it. Thank you

[Anat Tchetchik]

I think that the second specification is the right one for ardl(2,0,1) i.e. xtpmg D(1/2) x1 D.x2, ec(ec) lr(L.y x1 x2) replace pmg
As for your other question I think it that the lag structure affects the results of the entire model (and hence the long term coefficients) If the lag structure is miss-specified, the empirical results may be biased.