Is Prais Winsten method appropriate in this case? What is the difference from using xtregar / clustered standard error?

[Nana Lee]

Greetings.
I'm new to this forum, so I hope you could please kindly tell me if I break any rules somewhere.
Anyway, onward to the question.

For my assignment, I'm testing the impact of several macroeconomic aspects (GDP, foreign reserves, debt to GDP ratio, and exchange rate) to Credit Default Swap (CDS) spread.
The data is an unbalanced data panel with 4 countries (Indonesia, Malaysia, Philippine, and Thailand).
Total N is 59.
For each macroeconomic aspects, there are two independent variables used (one for the nominal itself, while the other one is the changes. For example, for GDP aspect, I make a variable for the nominal of GDP (log-transformed) and GDP growth). And thus, there are a total of 8 independent variables.

The models are homoscedastic (tested using White test) but are autocorrelated (tested using xtserial command).

And thus, here are the questions:
1) Is using Prais-Winsten in this case appropriate? I was also contemplating about using first difference / xtregar / xtreg with clustered standard error, but truthfully (and I'm kinda ashamed to admit this), but I don't really get the difference between all these methods.

2) Is Prais-Winsten a Random Effects / Fixed Effects / Pooled Least Square model? I tested using Chow & Hausman and found that PLS fits my needs the best.


Thank you for reading through this.
Best Regards.

[Carlo Lazzaro]

Nana.
welcome to the list.
The forum policy about classroom/home assignment is simple: please do not ask; please do not reply.
Usually, teachers/supervisors/prefessors are the main target-persons to whom those kind of questions should be addressed (basically because they are paid by students and/or their families/relatives to supevise and teach).
Things might be different if the original poster shows her/his personal engagement via posting what she/he typed and what Stata gave her/him back.
This step implies having read the FAQ (as requested).

[Nana Lee]

Nana.
welcome to the list.
The forum policy about classroom/home assignment is simple: please do not ask; please do not reply.
Usually, teachers/supervisors/prefessors are the main target-persons to whom those kind of questions should be addressed (basically because they are paid by students and/or their families/relatives to supevise and teach).
Things might be different if the original poster shows her/his personal engagement via posting what she/he typed and what Stata gave her/him back.
This step implies having read the FAQ (as requested).

I see. This is actually for my bachelor thesis and my supervisor is out of the country for at least two weeks. I would like to get some work done before he's back to the country. And I thought of posting results of the STATA, but I didn't think it would be necessary given the context of the question. But thanks for the heads up regarding rule.

[Carlo Lazzaro]

Nana:
thanks for providing further details.
Please, find below some remarks:
- if you really have 59 observations with 4 countries, any inference seems difficult to defend;
- your questions about methodological issues will require a relevant number of pages and worked-out examples (something that you can find in your supervisor's hand-outs/teachng notes, but not on a list);
- Stata .pdf manual (and the related references) is, as always, a very good first step to take to try to learn even complex statistical analyses/procedures.

[Nana Lee]

Nana:
thanks for providing further details.
Please, find below some remarks:
- if you really have 59 observations with 4 countries, any inference seems difficult to defend;
- your questions about methodological issues will require a relevant number of pages and worked-out examples (something that you can find in your supervisor's hand-outs/teachng notes, but not on a list);
- Stata .pdf manual (and the related references) is, as always, a very good first step to take to try to learn even complex statistical analyses/procedures.

Thanks for the kind reply.
Yeah, the data is very limited and that's actually already all the population I can find (via Bloomberg terminal).
And thank you for note, I will try to find more info regarding Prais Winsten.