Regression Analysis

[StlSluggers]

Mr. Pimp (or anyone else),

Maybe you can help me with some logic on this. run this regression analysis twice. The first time I ran it with two variables, but I realized that the first variable was actually the product of two other variables, so I separated it out into it's respective pieces and re-ran the numbers, which are what I posted in the first post.

Here are the R-squared measurements of some variables from each formula:

X1: 0.49799
Y: 0.31052
X1&Y: 0.90940

X2: 0.61531
Y: 0.31052
Z: 0.00363
X2&Y: 0.81803

X2 became more descriptive than X1, which makes sense because I pulled out the irrelevant part by segregating variable Z. However, the original combination of X1 and Y was actually more descriptive than X2 and Y in the new equation, and that makes absolutely no sense to me.

Can anyone explain how making one variable more descriptive actually made the combination of that variable and a constant less descriptive in the new formula?

[Big Pimpin]

It has to do with how the two variables are related. In this situation it may be that while X2 is the best predictor by itself, X2 and Z don't work as well together as X1 and Z.

It may be that Z is actually more related to X2 than it is to X1. So adding Z as a predictor with X2 doesn't actually add that much to the prediction. However, adding Z as a predictor along with X1 actually adds quite a bit to the analysis.

The whole thing is really confusing and hard to understand. It doesn't always make sense. Stats is a fickle, fickle being...

[StlSluggers]

The whole thing is really confusing and hard to understand. It doesn't always make sense. Stats is a fickle, fickle being...

I'm agree with that! Thanks for the help...