Can I use xttest3 for N=33 and t=8 (panel data)?

[Baby Marpaung]

Hello
I read the note the power of modified Wald statistic is very low in the context of fixed effects with "large N, small T" panels

Is it appropriate to use xttest3 in my data? (N=33 and t=8)
The result is
chi2 (33) = 432.18
Prob>chi2 = 0.0000
(Heteroskedasticity)

(Note: I use LM test for heteroskedasticity and get the result homoskedastic).

Thank you very much

[Carlo Lazzaro]

Baby:
welcome to the list.
You do not say which panel regression model your query refers to, nor report what you typed (please, see the FAQ #12).
This makes it difficult to get positive replies.

[Baby Marpaung]

Thank you for the reply.

I am using Stata for the first time. My reference to run my balanced panel data is Reyna, Oscar Torres. (2007). Panel Data Analysis Fixed and Random Effects using Stata (v. 4.2)

I am using xttest3 in Stata 13 to test heteroskedasticity

Modified Wald test for groupwise heteroskedasticity
in fixed effect regression model

H0: sigma(i)^2 = sigma^2 for all i

chi2 (33)  =    6358.28
Prob>chi2 =      0.0000

Here is the dataex

input int year byte provinsi float(lnp0 lngdrp1 lngdrp2 lngdrp3 lnpop) byte(_est_fixed _est_random)
2007 11  3.282789  30.36876  29.53626  30.25634 15.256246 1 1
2008 11  3.158276  30.23608 29.482777  30.28391 15.272706 1 1
2009 11   3.08191  30.04979  29.54405 30.227606 15.288785 1 1
2010 11  3.043394 30.068836  29.41609 30.384363 15.324708 1 1
2011 11  2.973932 30.110077 29.452663     30.44  15.34569 1 1
2012 11  2.968403 30.153303 29.487944  30.50045  15.36628 1 1
2013 11  2.867763  30.17632  29.49974  30.56413 15.386436 1 1
2014 11  2.893155  30.19933  29.51154 30.627813 15.406133 1 1
2007 12  2.631889  30.85329  31.16223 31.310495  16.36764 1 1
2008 12  2.529721  30.91212 30.984066  31.38962 16.383709 1 1
2009 12  2.443216 30.957853  31.13799  31.55024 16.399387 1 1
2010 12  2.425866  31.01341 31.285025  31.52599 16.382666 1 1
2011 12  2.427274 31.061325 31.221405 31.716225  16.39731 1 1
2012 12  2.367511  31.10621   31.2642  31.79497 16.411377 1 1
2013 12 2.3084102  31.14611  31.31096  31.86909 16.424868 1 1
2014 12 2.2383137  31.18601  31.35772  31.94321 16.437778 1 1
2007 13 2.4765384   27.5985   29.4604  30.13377 15.362605 1 1
2008 13 2.3674362  29.79043  29.52864 30.572456  15.37641 1 1
2009 13 2.2554934 29.924767 29.666016 30.519125 15.389943 1 1

I read the note the power of modified Wald statistic is very low in the context of fixed effects with "large N, small T" panels.
So, I want to ask: Is it appropriate to use xttest3 in my data? (N=33 and T=8)

Thank you very much

[Carlo Lazzaro]

Baby:
thnks for providing further details.
In your case -xttest3- seems powered enough to reject the null.
I would have been concerned about the limited power of the test if the (false) null was not rejected.