Building a predictive model - what to look for?

[tacomax]

That link does not even open up. Nice try though....

You need to enroll in How To Click On A Link 101 next semester. Nice try though.

Trust me....ANY variable you include will increase the R2. Go try it and tell me if it does not increase. That is a fact....

Adding a variable almost invariably increases R2. It doesn't always increase - I suggest you read the quoted passage from the book. R2 is monotonically increasing - but that doesn't mean it's strictly monotonic.

Since it's your word against someone who wrote a highly regarded book on econometrics, I suggest you take it up with the author and demand he make a correction.

[tacomax]

It opens up just fine for me. I'd been looking for a good online basic econometrics book. Thanks.

No problems. It's a good little book and isn't so confusing if maths isn't your strong point.

[BuddyBear]

You need to enroll in How To Click On A Link 101 next semester. Nice try though.



Adding a variable almost invariably increases R2. It doesn't always increase - I suggest you read the quoted passage from the book. R2 is monotonically increasing - but that doesn't mean it's strictly monotonic.

Since it's your word against someone who wrote a highly regarded book on econometrics, I suggest you take it up with the author and demand he make a correction.

Sorry, but it ALWAYS increases it. Trust me, i've taken enough stats coures to know at least the basic things of multiple regression

[Ganchrow]

Sorry, but it ALWAYS increases it.

The R² is only weakly increasing in the number of model regressors. There do exist pathological cases where the addition of a variable to a regression will have no effect on the coefficient of determination (R²). To wit:

What if the variable being introduced were collinear with an existing variable?

And as corollaries to the above:

What if every observation of the introduced variable were identical and there already existed a constant term?

With or without a constant term, what if the introduced variable were merely a column of 0s?

What if the R² of the regression were already 1?

[tacomax]

I see what you're doing there, Ganch. But BuddyBear has taken lots of stats courses - that seemingly trumps actual statistical theory and pure logic.