Help needed urgently on very simple question (I hope!)

[stats321]

Hi everyone,

I am regressing panel data, and I keep getting the error message near singular matrix when I change the Period function to fixed and I get the error message 'positive or non-negative argument to function estimate expected of RE innovation variance' when I change it to random.

Can someone please explain what this means in laymans terms and how I can hope to correct it?

I've removed variables to deal with collinearity, but no matter how many i remove, I still get these error signs.

PLEASEPLEASE help me!! I'm stuck and am losing hope as I am not really sure what I am doing

[EViews Gareth]

Could you post your workfile, and tell us the exact specification you're using?

[stats321]

I am trying to regress log(gdp) on log(airtransport) log(container port) log(electricity power loss) log(ICT imports) log(infant mortality) log(improved water source) log(raillines) (log road network) log(school enrolment) and Log(mortaility rates).

I have 57 cross sections spanning 1990-2009

I recognise that there are lots of variables and so I cut them, but regardless of which ones I cut out, or even if I just estimate using one dependent variable, it still gives me those error messages.

Am I missing something?

[startz]

It would be much easier to track down what you've done if you posted and EViews workfile, including the equation you are trying to estimate.

[stats321]

Sorry I am new to this.

In an ideal world, I would like to estimate:

log(gdp)= log(elecloss) log(airfreight) log(ictimport) log(infmort) log(mobile) log(port) log(railtot) log(roadtot) log(secondsch) log(water)

In panel options, it's not letting me change for fixed/random effect for the period.

I have never worked with panel data before so I could be missing something glaringly obvious and stupid!

log3.wf1

(138.68 KiB) Downloaded 219 times

[EViews Gareth]

Your dependent variable (GDP) has no cross-section variance. i.e. it is the same for all cross-sections. You can't use the panel options if that's the case.

[stats321]

In what sense is it the same as I have (hopefully) got a different GDP value for each corresponding year for each country?

[startz]

Your values for electric power are the same for every country, at least for 1990 and 1991. Therefore you can't estimate a period fixed effect.

Most likely, you have a data error.

[stats321]

Sorry for being such a novice at these things! I am trying to teach myself and it is not going so well.

[EViews Gareth]

In what sense is it the same as I have (hopefully) got a different GDP value for each corresponding year for each country?

Take a look at the values for GDP in 1990 for cross-section 1. It is identical for GDP in 1990 for cross-section 2, and cross-section 3 etc....

As Startz pointed out, you have data errors.

[stats321]

I think that is working now- I knew it was something stupid on my part.

Thank you so much for your help I really appreciate it and I am sorry if I bother you with more silly questions!