Not displaying the F-Statistics and and Prob value when using robust command

[Mahadevan Suthakar]

Dear Statalist,

I am having an balanced panel data set (banks profitability determinants factors) and using Fixed Effects with time dummy (after using testparm command). I run the multiple regression using streg command as follows,

.xtreg ROAA SIZE CCAR LATA DPTA CRCO COIN REVD i.Year, fe

Outcome is ok with F(13,47) = 49.77, and Prob > F = 0.0000. However, I added robust after fe to control heteroscedasticity.

.xtreg ROAA SIZE CCAR LATA DPTA CRCO COIN REVD i.Year, fe

The outcome does not indicate the F-Statistics and and Prob value. Refer the outcome as follows,

Fixed-effects (within) regression Number of obs = 70
Group variable: ComCode Number of groups = 10

R-sq: within = 0.9323 Obs per group: min = 7
between = 0.5450 avg = 7.0
overall = 0.5168 max = 7

F(9,9) = .
corr(u_i, Xb) = -0.8828 Prob > F = .

(Std. Err. adjusted for 10 clusters in ComCode)
------------------------------------------------------------------------------
| Robust
ROAA | Coef. Std. Err. t P>|t| [95% Conf. Interval]
-------------+----------------------------------------------------------------
SIZE | .9354233 .1792884 5.22 0.001 .5298448 1.341002
CCAR | .0338517 .0135996 2.49 0.034 .0030872 .0646162
LATA | .0101302 .0117428 0.86 0.411 -.0164338 .0366942
DPTA | .0046298 .0092143 0.50 0.627 -.0162145 .025474
CRCO | -.5315502 .0427412 -12.44 0.000 -.6282376 -.4348629
COIN | -.0661699 .0093161 -7.10 0.000 -.0872443 -.0450955
REVD | .5407771 .0805341 6.71 0.000 .3585963 .7229578
|
Year |
2012 | -.2612143 .0972922 -2.68 0.025 -.4813045 -.0411242
2013 | -.4931095 .1163411 -4.24 0.002 -.7562913 -.2299277
2014 | -.8438651 .1134289 -7.44 0.000 -1.100459 -.5872712
2015 | -.971674 .1276638 -7.61 0.000 -1.26047 -.6828784
2016 | -1.201661 .1439492 -8.35 0.000 -1.527296 -.8760251
2017 | -1.502331 .2238174 -6.71 0.000 -2.008641 -.9960207
|
_cons | -14.11403 4.528691 -3.12 0.012 -24.35864 -3.86942
-------------+----------------------------------------------------------------
sigma_u | 1.0880954
sigma_e | .17499072
rho | .97478806 (fraction of variance due to u_i)
------------------------------------------------------------------------------

Why I am not getting F-Statistics and and Prob value after using robust command ?

[Carlo Lazzaro]

Mahadevan:
welcome to this forum.
What you report is pretty common: see -help j_robustsingular-.

[Mahadevan Suthakar]

Thanks a lot Sir for your kind assistance. Could you please explain further ? Is there any other commands to be use to resolve the issue ? [see -help j_robustsingular-.] - where can i check ?

[Carlo Lazzaro]

Mahadevan:
read the paragraph Are you using a svy estimator or did you specify the vce(cluster clustvar) option? under -j_robustsingular-.
In brief, there's nothing you can do to avoid Stata omitting F-test value and related p-value if you invoke cluster-robust standard errors.
As an aside, please call me Carlo, like all on (and many more off) the list do. Thanks.

[William Lisowski]

I note in post #3

[see -help j_robustsingular-.] - where can i check ?

and I would like to be sure it is clear that post #2 suggests typing the command

help j_robustsingular

and reviewing the output from that command,

I mention this in part because I initially believed I had misread post #2 - how (in the absence of Carlo's advice!) would I have ever thought to type "j_robustsingular" as an argument to help? And having done so, I notice that the output from help doesn't link to anywhere else to give it context.

My thanks to Carlo for supplying a useful reference I would never have found on my own.

[Mahadevan Suthakar]

Dear Sir,

Sorry for troubling again. Used both

(1) fe vce(robust)
(2) fe vce(cluster ComCode )

options.

Under both options no F-test value and related p-value. Is that possible to proceed with out F-test value and related p-value?

Data set also enclosed.

Kind regards,
Mahadevan

[Carlo Lazzaro]

Mahadevan:
what you report is expected, as -robust-and -cluster- option do the very same job.
The missing F-test is, in general, a minor nuisance: hence, you can proceed.
As an aside, please read the FAQ about how to share what you typed and what Stata gave you back (attachments, let alone spreadsheet that no one on this list will ever download to the risk of active contents, are discouraged). Thanks.

[Mahadevan Suthakar]

Dear Sir,

Thank you very much for your kind assistance.

kind regards,
Mahadevan

[Carlo Lazzaro]

Mahadevan:
please call me Carlo, like all on (and many more off) the list do. Thanks.

[Mahadevan Suthakar]

Dear Carlo,

Sorry for troubling again. Further to the above, rather than using 'xtreg' command, if i use 'areg' command with the 'robust' to control heteroscedasticity i am getting the F-test value and related p-value

The Command i used is as follows,

areg ROAA SIZE CCAR LATA DPTA CRCO COIN REVD i.Year, absorb( ComCode) vce(robust)


The Outcome is

Linear regression, absorbing indicators Number of obs = 70
F( 13, 47) = 47.14
Prob > F = 0.0000
R-squared = 0.9611
Adj R-squared = 0.9429
Root MSE = 0.1750

------------------------------------------------------------------------------
| Robust
ROAA | Coef. Std. Err. t P>|t| [95% Conf. Interval]
-------------+----------------------------------------------------------------
SIZE | .9354233 .4072873 2.30 0.026 .1160669 1.75478
CCAR | .0338517 .0108308 3.13 0.003 .012063 .0556404
LATA | .0101302 .0099553 1.02 0.314 -.0098972 .0301576
DPTA | .0046298 .0094105 0.49 0.625 -.0143017 .0235612
CRCO | -.5315502 .0661518 -8.04 0.000 -.6646304 -.39847
COIN | -.0661699 .0078957 -8.38 0.000 -.082054 -.0502858
REVD | .5407771 .0651079 8.31 0.000 .4097969 .6717572
|
Year |
2012 | -.2612143 .1427715 -1.83 0.074 -.5484335 .0260048
2013 | -.4931095 .2057801 -2.40 0.021 -.9070857 -.0791332
2014 | -.8438651 .2338337 -3.61 0.001 -1.314278 -.3734524
2015 | -.971674 .2994713 -3.24 0.002 -1.574133 -.3692153
2016 | -1.201661 .3670915 -3.27 0.002 -1.940154 -.4631679
2017 | -1.502331 .4607756 -3.26 0.002 -2.429292 -.5753699
|
_cons | -14.11403 8.533795 -1.65 0.105 -31.28181 3.053751
-------------+----------------------------------------------------------------
ComCode | absorbed (10 categories)


The Coefficient and the outcome are same with 'xtreg' command however only small change in 'P>|t|' and anyway that is not changing the decision on the significant level.

My Question is

Can I use 'areg' [areg ROAA SIZE CCAR LATA DPTA CRCO COIN REVD i.Year, absorb( ComCode) vce(robust)] with the robust option?

Kind regards,
Suthakar

[Carlo Lazzaro]

Muhadevan:
the following technical note in -areg- entry, Stata .pdf manual warns:

Although the point estimates produced by areg and xtreg, fe are the same, the estimated VCEs
differ when vce(cluster clustvar) is specified because the commands make different assumptions about whether the number of groups increases with the sample size.

Hence, what you got from -areg- is possibly not what you supposed it to be.

[Mahadevan Suthakar]

Thank you Carlo. Few more clarifications,

(a) In what circumstances -areg- command is used in Panel Fixed Effects?

(b) Is unit root / stationarity test required for Panel Data analysis (Panel with 7 Years) ? What is the stata command for unit root test?

(c) Is Serial correlation a problem for micro panel with 7 years data. What is the Command in stata to check serial correlation ?

Regards,
Mahadevan

[Carlo Lazzaro]

Mahadevan:
(1) when the number of groups is not expected to increase as sample get larger (see -help areg-);
(2) see -help xtunitroot- (usually relevant for T>N panel dataset);
(3) see -search xtserial- (community-contributed programme). Serial correlation (as well as heteroskedasticity) can be accounted for via cluster robust standard errors.

[Mahadevan Suthakar]

Many thanks Carlo.

(2) Can you exactly post 'T' and 'N' stand for what ? and

(3) Can you post the stata command for 'cluster robust standard errors' ?

Kind Regards,
Mahadevan

[Carlo Lazzaro]

Mahadevan:
1) T=time-series deimension; N=cross-sectional dimension
2) -xtreg, fe vce(cluster panelid)- or -xtreg, vce(cluster panelid)-