Comparing output between different values of a variable

[rbbtchnnl]

Hello, I hope this is in the correct forum. I'm using eviews 7 to do a regression on cross sectional data but I need to know how to compare outputs between different values for a variable. For example, if I'm regressing income against age, I want to examine the effects of age=25 and age=30. How exactly do I do this in eviews? I've tried using a sample with IF AGE=25 but it produces a near singular matrix. Someone else suggested using age(25) as a variable but I'm not sure what that does to the data.

Can anyone help please?

[EViews Gareth]

If you only want to regress using those observations where age=25, then setting the sample to "if age=25" is the correct way to do it. The fact you get a singular matrix, probably means you've got perfect collinearity for that sample.

[rbbtchnnl]

The thing is, regressing it against age produces a regression with no issues. It's only when I set age to something, say age=25, that it says it's a singular matrix...

[rbbtchnnl]

Oh wait!

Question, if I'm regressing income against age, education, and experience, for example, I'd use

ls inc age edu exp c.

But then if I set age=25, then I'd use the sample and go with IF age=25 and run the regression without the age variable as

ls inc edu exp c, right?

[EViews Gareth]

Indeed.

[startz]

Run the full model. Suppose the coefficient on age is 7. Then the regression is telling you that the difference in income between ages 25 and 30 is
7*(30-25)

[rbbtchnnl]

Thanks guys!

One more question.

I'm trying to run a White Test on a series with 10,000 obs. The chi squared is much, much higher than obs*r-squared and my professors don't seem to have used that method. They just mention that the result of an example has "low power" indicating heteroskedasticity. I'm unsure what statistic they are looking at to determine this. Can anyone help?