Fixed effects in panel model

For econometric discussions not necessarily related to EViews.

Moderators: EViews Gareth, EViews Moderator

Ydwolf
Posts: 14
Joined: Thu Oct 06, 2011 11:42 pm

Fixed effects in panel model

Postby Ydwolf » Thu Apr 12, 2012 6:50 am

Dear,

I have a question which is probably very basic for you guys.

I'm working with panel data (see workfile) and now I want to run very basic OLS regressions with them, like for example ars c trade (=transmission1_trade in the workfile). I sent my first results to my professor and he said: "be careful when estimating a panel model, because there are fixed effects in the panel variable".

Can somebody please explain this to me in human language? In the case of "transmission1_trade" for example, both ars and trade are panel variables, right? And how can I test for and interpret these so-called "fixed effects". What is its effect on the interpretation of the model?

I'm so sorry for this stupid question, but I can't seem to find my answer on the internet...

Tnx!
Attachments
diffusiekanalen12042012.wf1
(31.39 KiB) Downloaded 127 times

Ydwolf
Posts: 14
Joined: Thu Oct 06, 2011 11:42 pm

Re: Fixed effects in panel model

Postby Ydwolf » Thu Apr 12, 2012 7:14 am

To avoid any confusion, I added my original excel file too.
In the meantime, I figured out that I should adjust something in "panel options" when I estimate my equation. Both "cross-sections" and "periods" are set by default on "none". Should I adjust both to fixed? Again, sorry, but I do not seem to understand the manual...
Attachments
proxies_2004_2005_.xlsx
(36.16 KiB) Downloaded 115 times

Ydwolf
Posts: 14
Joined: Thu Oct 06, 2011 11:42 pm

Re: Fixed effects in panel model

Postby Ydwolf » Thu Apr 12, 2012 8:04 am

Okay, I should first run a Hausman test on random effects?

But when I try doing that, after setting both cross-sections and periods to "random" and then run the Hausman test. I get the following error message:

"near singular matrix in computation of Swamy-Arora RE component variances for estimate of RE innovation variance"

What am I doing wrong?

Tnx!

Ydwolf
Posts: 14
Joined: Thu Oct 06, 2011 11:42 pm

Re: Fixed effects in panel model

Postby Ydwolf » Fri Apr 13, 2012 1:02 am

I'm sorry to disturb you guys once more (I swear, this is the last time).
In the meantime, I found an answer to most of my questions. For all the simple equations I liked to run, I did the following:

1) Specify my model (in dit geval ars c trade)
2) Select with fixed effect on cross section & period
3) View --> Fixed/Random Effect Testing --> Redundant Fixed Effect - Likelihood Test

For all my regressions (except for 1), there seem to be significant fixed effects.

Then, I wanted to test the random effects model against the fixed:

1) Specify my model
2) Select with random effect on cross section & period
3) View --> Fixed/Random Effect Testing --> Correlated Random Effects - Hausman Test

But then I often get the error "near singular matrix"...

1) My question now is, can I just not worry about this and run fixed models for all my regressions?
2) And last question, for the one regression where fixed effects are not significant (ars c finaspain), I did the Hausman test, and I get the error: Two-way random effects not allowed with unbalanced data" (probably because there are some missing values). What should I do with this equation?

You can find all of my data and tests in the workfile.

Can somebody please respond because it is extremely urgent. I promise, I won't bother you any further after this.

Thank you so much

Yasmin
Attachments
diffusiekanalen12042012.wf1
(129.46 KiB) Downloaded 166 times


Return to “Econometric Discussions”

Who is online

Users browsing this forum: No registered users and 3 guests