Insufficient number of observations

[jasperq]

Hi,
I am having a problem when I want to estimate my equation. I have a set of paneldata (20 countries, 8 years) on a certain variable that is available for all countries over all years. Now I want to specify my explanatory variable in five levels, so I created five new variables. Now I want to make a regression but eviews tells me I have an insufficient number of observations.
Does anyone kwow if there is an easier way to split my explanatory variable into five new variables so that eviews will not make a problem. I attached my data in an excel format.

[startz]

It would be much better if you
(a) posted the EViews workfile you actually used.
(b) showed exactly what regression you ran.

[jasperq]

Okay, I have my workfile attached. As you can see I have been struggling with discriminating unden for the different levels. What I would like to do is to run a regression that includes an explanatory variable that combines unden and level, ie: I would like to see the effect each rate of unden at a particular level has on my explained variable. When I try dummy variables, not only I obtain different coefficients if I run the regression for one particular level but obviously I cannot incorporate all dummy variables at once. Do you have any idea how I could fix this and still see the different effects? I would really appreciate your help!

Eventually I want my equation to look like this: avg_dur c "unden at level1" "unden at level2" "unden at level 3" "unden at level 4" ... and the other explanatory variables.

[EViews Gareth]

It would be much better if you
(b) showed exactly what regression you ran.

[jasperq]

This is the equation I wanted estimated: avg_dur c density1 density2 density3 density4 nrr epl realminwag.
I can't post the regression's outcome, since eviews gives an error. In this first regression I wanted to run I used an OLS model without any fixed effects.
Is there a better way to show what regression I want to run? Thanks for the reply!

[startz]

This is the equation I wanted estimated: avg_dur c density1 density2 density3 density4 nrr epl realminwag.
I can't post the regression's outcome, since eviews gives an error. In this first regression I wanted to run I used an OLS model without any fixed effects.
Is there a better way to show what regression I want to run? Thanks for the reply!

Your workfile doesn't include the variables density1 density2 density3 density4

You might want to make an equation object that has the exact regression you want to run and include that in the workfile you post.

[jasperq]

Ah I see, this one does.

I can't add an equation to my workfile as eviews will not let me run the regression. I think I did not understand you correctly.

[startz]

You can add an equation by typing

Code: Select all

equation eq01

and then double-clicking on eq01 and filling out the specification.

But in any event, your problem is that for every observation at least one of density1 density2 density3 density4 is NA. You have zero complete observations.

[jasperq]

Okay, thanks for your reply. I still can't seem to manage to estimate the equation. I know how to do this if there is no error given by eviews but I still can't specify it here.

Have you got an idea on how I would still be able to see the effect of the density at different values for "level"?

[startz]

EViews is doing what it's supposed to. There is no way to estimate an equation without any data. You need to figure out why your variables are missing so much data.

[jasperq]

Thanks for the replies, I chose to exclude some variables on a theoretical basis and this at lest allowed me to run the regression.