How do you find the r-squared and f-stat for an ARDL panel data model?

[Stephen Devine]

Hi, I'm running a panel data ARDL model to see If there's a relationship between GDP per capita and female secondary school enrolment, including infant mortality and female labour force participation rate as additional dependent variables.

My code is:

xtpmg d.lngdppc d2.secen d.lfpr d2.lninfmort , lr(l.lngdppc secen lfpr lninfmort) dfe replace ec(ec) cluster(country).

If it is useful, N=3 and T=58, I used the Fisher test for unit root.

I have two questions.
Firstly, is it acceptable for me to have I(2) series in the model? From the reading I have done (Persaran, 2001) I have only seen I(0) and I(1) mentioned.
Secondly, how do I obtain the r-squared and f-stat values for the model?

Thanks in advance for any help.

[Stephen Devine]

In addition, I'm unsure if I can have I(1) series and I(2) series, both are in the code above

[Philipp Allroggen]

Hi Stephen, as far as I know only I(0) and I(1) are feasible in ARDL.
Unfortunately, I do not know how to get the r-squared and f-stat. Maybe if you use the option -full-. But I am far from beeing an expert, I am just getting started with xtpmg myself.

Personally, I would suggest to run extra unitroot tests im possible. Other tests might indicate that your data is stationary I(1).
Also, I am not sure, if Pooled Mean group estimator is best suited for your type of Data as you only have N=3. Maybe you can think about regular FE Estimator, because with T=58 you could argue that the bias diminishes.

Ps: use code delimitors when you post a code. That indicates that you read the FAQ and it might increase the chances a real expert might come across and helps you out