Selection of appropriate method in cross-sectional dependence (fe, re, pols?)

[Farheen Akram]

Salam and Good Evening to all,
Please help me in running cross sectionally dependent random effect model in panel data analysis. As the result of Hausman test favors RE, but unfortunately, the command for random effect is not running in STATA with cross-sectional dependence.
The command used is [xtscc DV IV1 IV2 ...... IVn, re]. The question is:
1. If we have to follow Hausman selection of appropriate method?
2. If there is any other command in cross sectional dependence to run RE?
3. If running with POLS, how to handle the problems of heteroskedasticity, serial correlation and cross-sectional dependence?


Comand used.
hausman FE REgls, sigmamore

Here, Rendom effect is appropriate.
But when i entered the comand
xtscc DV IV1 IV2 ...... IVn , re
Option re not allowed.

Please give me your best suggestions

[Carlo Lazzaro]

Farheen:
you do not say whethet your panel is a large N, small T one (or the other way round).

[Marcos Almeida]

I wish to remark (as recommended in the FAQ), that - xtscc - is in fact a user-written program aimed at performing "regression with Driscoll-Kraay standard errors", whose author is Daniel Hoechle.

According to the help files (you may read here), this program wouldn't allow RE, but only FE and pooled regression.

[Farheen Akram]

Farheen:
you do not say whethet your panel is a large N, small T one (or the other way round).

My data is unbalanced panel data. With 7 years and 188 observations.

[Farheen Akram]

I wish to remark (as recommended in the FAQ), that - xtscc - is in fact a user-written program aimed at performing "regression with Driscoll-Kraay standard errors", whose author is Daniel Hoechle.

According to the help files (you may read here), this program wouldn't allow RE, but only FE and pooled regression.

Thank you so much for sharing your knowledge sir.
In my case the hausman test did recomend the fe. As the p-value is greater than the 0.05.
Should i use pooled regression?

[Marcos Almeida]

Sorry, but the theme (regression with Driscoll-Kraay standard errors) is not at all one I feel confident to discuss.

This being said, I wish to underline that:

The above-mentioned user-written program doesn't allow RE option.

I suspect there must be some reason, and you should search on this.

In other words, I fear you need to reflect about the model itself, before making a final decision. It seems to me surprising the you wish to perform a RE model (dutifully based on the Hausman test), albeit selecting a type of regression which, as it seems to be, wouldn't allow for that.

To end, when the pattern of data is longitudinal, a panel data analysis (instead of pooled regression) tends to be a wise bet, so to speak.

Hopefully you will get further advice, ideally from members with expertise in this special kind of regression.

[Carlo Lazzaro]

Farheen:
since you have a large N, small T panel dataset, I would go -xtreg, re- with robustified/clustered standard errors (the do the same jo under -xtreg-).

[River Huang]

If you are concerned about cross-sectional dependence, you can try (ssc install) -xtcce- command: Common Correlated Effects Estimation for Static/Dynamic Panels with Cross-Sectional Dependence.

[Farheen Akram]

If you are concerned about cross-sectional dependence, you can try (ssc install) -xtcce- command: Common Correlated Effects Estimation for Static/Dynamic Panels with Cross-Sectional Dependence.

Thanks for sharing your knowledge.
Would you like to suggest ke how to deal with heteroskedasticy ?

[River Huang]

Thanks for sharing your knowledge.
Would you like to suggest ke how to deal with heteroskedasticy ?

Well, I didn't find an option for -xtcce- command for adjusting heteroskedasticity either. You might want to contact the author for this possibility.

[Farheen Akram]

xtcce comand give me the following error. whtadoes it means?

[elif dasci]

what is the code, please?:
Pesaran, Ullah and Yamagata (2008) Cross Section Dependence Test

[Nick Cox]

elif:

Please

1. don't post new questions in existing threads

2. don't post the same question in two or more threads

https://www.statalist.org/forums/help#adviceextras #1 gives advice on these

3. do give references in full form

https://www.statalist.org/forums/help#references

4. do use search as recommended.

https://www.statalist.org/forums/help#before

In this case,

search xtcsi

gets you there (after a search using your keywords).

[elif dasci]

elif:

Please

1. don't post new questions in existing threads

2. don't post the same question in two or more threads

https://www.statalist.org/forums/help#adviceextras #1 gives advice on these

3. do give references in full form

https://www.statalist.org/forums/help#references

4. do use search as recommended.

https://www.statalist.org/forums/help#before

In this case,

search xtcsi

gets you there (after a search using your keywords).

I will consider them. Thank you for your warnings.

[Nicu Sprincean]

If you are concerned about cross-sectional dependence, you can try (ssc install) -xtcce- command: Common Correlated Effects Estimation for Static/Dynamic Panels with Cross-Sectional Dependence.

@River Huang: in the help file of the xtcce command, one says that this method is suitable for medium to large T and N. In my case, for instance, I have a micro panel, with small T (16) and large N (over 250). What method could be suitable in this case? Thanks.