F test for Panel Data

[Imran]

Dear,
I want to estimate the Ftest for the panel data. How i Estimate it.

[Carlo Lazzaro]

Imran (as per FAQ #6, please take a look at how to re-register as suggested by the requirements of this forum. Thanks):
as your question is way too broad, posting more details will help you in getting helpful replies.
I don't understand which F test you have in mind, but please note that -xtreg- produces an F-test by default:

use http://www.stata-press.com/data/r13/nlswork 
xtreg ln_w grade age c.age#c.age ttl_exp c.ttl_exp#c.ttl_exp tenure c.tenure#c.tenure 2.race not_smsa south, fe

[Imran]

Thanx Carlo for Suggestion.
Acually i am working on Panel Data.According to my Knowledge, In Panal Data F test is use to check either Fixed effet is better then Pooled Ols. I dont know how to apply F test or what is command to apply test and i also apply Hausman Test to examine fixed effect is better than or Random effect. I hope you understand my question.

Kind regards, Imran

[Carlo Lazzaro]

Imran (as per FAQ #6, please take a look at how to re-register as suggested by the requirements of this forum. Thanks):
now I see the issue.
If you run the Stata code I provided you with in my previous post, the F-test automatically reported as a footnote in result table tells you that -xtreg, fe- is actually better than pooled -ols-.
For -hausman- specification test, expanding on the previous Stata code, you may want to try:

use http://www.stata-press.com/data/r13/nlswork 
xtreg ln_w grade age c.age#c.age ttl_exp c.ttl_exp#c.ttl_exp tenure c.tenure#c.tenure 2.race not_smsa south, fe 
estimates store fe
xtreg ln_w grade age c.age#c.age ttl_exp c.ttl_exp#c.ttl_exp tenure c.tenure#c.tenure 2.race not_smsa south, re
estimates store re
hausman fe re

I would also reccommend you to take a look at -help xtreg- and -help hausman- and related entries in Stata 13.1 .pdf manual.

[Imran]

hausman fixed random

---- Coefficients ----
| (b) (B) (b-B) sqrt(diag(V_b-V_B))
| fixed random Difference S.E.
-------------+----------------------------------------------------------------
x| -.0001601 -.0001904 .0000303 .0000295
y| -.0104284 .0008819 -.0113103 .0038792
z| -.0348392 -.0282139 -.0066254 .0053239
------------------------------------------------------------------------------
b = consistent under Ho and Ha; obtained from xtreg
B = inconsistent under Ha, efficient under Ho; obtained from xtreg

Test: Ho: difference in coefficients not systematic

chi2(3) = (b-B)'[(V_b-V_B)^(-1)](b-B)
= 12.61
Prob>chi2 = 0.0056
I apply Hausmen test. I did not understand the result. so what the result indicate either what we choose fixed or random

[Marcos Almeida]

Hello "Imran",

Just type "help hausman", then click on "[R] hausman" (Stata Base Reference Manual) and go to the first example on page 769-771.

[Carlo Lazzaro]

Imran (as per FAQ #6, please take a look at how to re-register as suggested by the requirements of this forum. Thanks):
Seconding in full Marcos' guidance, please consider that your query is also covered in any decent econometrics textbook.
Anyway, the result of the -hausman- specification test points you out toward -xtreg, fe-.

[Imran]

thx Marcos Almeida
Please tell me about the F test. Acually i am working on Panel Data.According to my Knowledge, In Panal Data F test is use to check either Fixed effet is better then Pooled Ols. I dont know how to apply it,
I hope you understand my question.

[Imran]

Dear Carlo,
i did not understand about your suggestion. It is first experience of my life to use such type of forum. I hope you understand my problem

[Carlo Lazzaro]

Imran (what between brackets still applies, Thanks)
you have already received the replies you were searching for by Marcos and myself.
Now the question is: why you seemingly don't like the replies you got and are asking for more?
Please, note that this has nothing to do with being unfamiliar with using Statalist.
Just to wrap up (bold intended)
#4:

...If you run the Stata code I provided you with in my previous post, the F-test automatically reported as a footnote in result table tells you that -xtreg, fe- is actually better than pooled -ols-....

;
#6:

Just type "help hausman", then click on "[R] hausman" (Stata Base Reference Manual) and go to the first example on page 769-771.

;
#7

the result of the -hausman- specification test points you out toward -xtreg, fe-.

;

Please, once more:
- take a look -help xtreg- and -help hausman- and related entries in Stata 13.1 .pdf manual;
- skim through any decent econometrics textbook.

[Nick Cox]

Everyone has to start at the beginning, which is why we provide Advice in the FAQ.

This applies here too:

4. Elementary questions?

Beginners' questions are not out of order, but very basic or elementary questions that you should be able to answer for yourself will get little or no reply. There is an underlying expectation that you have tried to read the documentation or basic literature at your level. So, Statalist is not for questions on how to do regression in Stata, the difference between means and medians, or even what instrumental variables are. If you are in doubt about where to draw the line, look at previous questions. If you get this wrong, you are unlikely to get flamed. You just may not get an answer.

[Marcos Almeida]

Hi, Imran,

As Carlo underlined, please re-register with you name and family name (just click on the "contact us" button below to the right and give this information. Meanwhile, until the admnistration perform the changes demanded by you, please use your name and family name inside your message.

Also as Carlo underlined, you gave the impression your question was mostly related to the background knowledge of panel data, and that you can easily get just by reading - now quoting Carlo - "decent textbooks".

Or the Stata Manual.

If - as I understood - you are starting to delve into longitudinal studies, please don't underestimate the Stata Manual, There you will have most of the background information you're longing for.

Just to give you an example, and going directly to the F-tests under xtreg, please type "help streg", then click on "[XT] xtreg" on the top of the page. Now, you'll be "in the realms" of the Stata Longitudinal-Data Panel-Data Reference Manual.

Flicking through pages, I gather on page 370 you'll get the answer for you question:

The F statistic tests that the coefficients on the regressors [...] are all jointly zero.

Hopefully that helps!

Best,

Marcos

[Nima Din]

wrong post, sorry!

[Ridwan Sheikh]

Dear Carlo Lazzaro .
I have time-constant dummies as my explanatory variables along with other covariates, the 'fe' estimator will drop them in de-meaning. Is it OK to run a random effect model and report wald-chi2 instead of F-statistic as test for joint significance. My wald chi2 is 3394.17 with a p-value of 0.0000. Can i infer, all the explanatory variable are significant predictor of outcome (dep. variable) .
I am running a first stage regression of endogeneous variable on the set of instruments and other covariates in the model, I am running something like:

xtreg endogeneous instruments covariates, cluster (pair)

this is my crude way of checking instrument relevance, whether the regressors (including instruments) are jointly determining y (endogeneous variable). Is wald-chi2 valid here. The instruments are time-constant, so fe estimator would drop them, therefore i am not using fixed effects estimator, rather testing the instrument relevance using random effects model and reporting wald-chi2 instead of F-statistic. Is this a correct approach ? Jeff Wooldridge
Thanks,
(Ridwan)

[Carlo Lazzaro]

Ridwan:
the main issue here seems to rest on the appropriateness of the -fe- vs the -re- estimator as far as your dataset is concerned.
What does the -hausman- (or the community-contributed module -xtoverid-) tell you?