Panel data and autocorrelation

[David Junius]

Do panel data need autocorrelation?

[Carlo Lazzaro]

David:
welcome to this forum.
Autocorrelation should be taken into account via cluster-robust standard error (see options -cluster()- and -robust- under -xtreg-).

[David Junius]

Thankyou Carlo, I'm sorry if what I asked maybe unclear or I don't attached data needed for you to answer my question since I'm literally new here and am still a undergraduate student. Can you please show me where is this "options" in the toolbar? I'm using Stata/SE 13.0. I am also in need of help since my data suffers from heteroscedasticity and autocorrelation, for the hetero my college mates told me to use robust to solve the problem but I can't find any fit solution for the autocorrelation problem because what I found on the internet mostly say that autocorrelation problem is affiliated with time series data. Maybe I'll try to attach my data here (sorry if I do it wrong)

. reg roa esg size lev logage ind cntry1

Source | SS df MS Number of obs = 1302
-------------+------------------------------ F( 6, 1295) = 33.43
Model | 8000.16012 6 1333.36002 Prob > F = 0.0000
Residual | 51649.0846 1295 39.883463 R-squared = 0.1341
-------------+------------------------------ Adj R-squared = 0.1301
Total | 59649.2447 1301 45.8487661 Root MSE = 6.3153

------------------------------------------------------------------------------
roa | Coef. Std. Err. t P>|t| [95% Conf. Interval]
-------------+----------------------------------------------------------------
esg | .0317913 .013604 2.34 0.020 .005103 .0584796
size | -.295232 .0745342 -3.96 0.000 -.4414529 -.149011
lev | -.1075847 .0107117 -10.04 0.000 -.1285989 -.0865705
logage | -.758627 .6035218 -1.26 0.209 -1.942615 .4253606
ind | -3.250453 .4309578 -7.54 0.000 -4.095905 -2.405001
cntry1 | 1.42577 .6239347 2.29 0.022 .201736 2.649803
_cons | 13.26432 1.141113 11.62 0.000 11.02569 15.50296
------------------------------------------------------------------------------

. hettest

Breusch-Pagan / Cook-Weisberg test for heteroskedasticity
Ho: Constant variance
Variables: fitted values of roa

chi2(1) = 198.29
Prob > chi2 = 0.0000

. xtserial roa esg size lev logage ind cntry1

Wooldridge test for autocorrelation in panel data
H0: no first-order autocorrelation
F( 1, 270) = 2.515
Prob > F = 0.1140

[Carlo Lazzaro]

David:
-please do not share attachments (see the FAQ on this and other posting-related topics). By the way, please note that, due to the risk of active contents, nobody on this list would ever download spreadsheets from unknown sources. Use -dataex- instead;
-please use CODE delimiters

these are CODE delimiters

to post what you typed and what Stata gave you back;
-you have an apparent heteroskedasticity issue in your regression model: -robust- option from -regress- will deal with it;
- being devised for panel data analysis, the community-contributed command -xtserial- (as per FAQ, please note that, whenever you use a non-official Stata command, you should highlight it) does not support -regress-. Use -estat vce, corr-, instead to detect quasi-extreme multicolinearity issues, instead.
As an aside, I would recommend you to rehearse your knowledge of panel data analysis via any decent (panel data) econometrics textbook. Stata users like https://www.stata.com/bookstore/micr...metrics-stata/.

[David Junius]

Thanks Carlo Lazzaro for the suggestion, I'll try to execute it asap. I have one more question if you are willing to answer, what is the best command for heteroscedasticity, multicollinearity and autocorrelation for Fixed Effect Model? I just found out that I am supposed to use FE instead of OLS/PLS and I think I ran the test wrong. Thanks!

[Carlo Lazzaro]

David:
1) testing for heteroskedasticity: see https://economics.stackexchange.com/...s-time-series;
2) testing for multicollinearity: see outcome from -estat vce, corr-;
3) testing for autocorrelation: see the user-written command -xtserial-.

[David Junius]

Carlo Lazzaro Thankyou so much carlo and I'm sorry if it's hard for me to understand, hope you still willing to answer my questions in the future.

Regards.

[Carlo Lazzaro]

David:
we're all have been (and, as far as I'm concerned; I often find myself being still a) newbie(s).
The most prolific contributor to this forum - Nick Cox - once wrote (surely in an English better than mine): "We are all beginners: some of us are only more experienced".
I find in Nick's sentence the essence of the lifelong learning.
Due to its muldisciplinary flavour, the Stata forum is a valuable opportunity to learn about Stata and statistics at large.
Everybody's posts are welcome: the only request is reading and acting on the FAQ.