I know we have some stat heads here, and I need some help refreshing my memory on regression analysis. I have an output (N = 14) that I think might be based on three variables. When I ran the analysis, I got the following figures:

R-square = .86886
Intercept = -12.92890
Beta X = 13.1782
Beta Y = 0.80400
Beta Z = 16.5652

The Beta for Z seemed extraordinarily high to me, so I ran the analysis with just Z, and I got a number that seemed really odd to me:

R-square = 0.00363

So I ran the analysis again with X and Y, but without Z, and got:

R-square = 0.81803

Does this mean that variable Z is really doing nothing for my overall equation?

That's exactly what it means. An R-sq of 1 would mean that 100% of the variability in your dependent variable can be explained by your independent variable. With your R-sq of almost zero, it means Z is adding virtually nothing.

Big Pimpin wrote:That's exactly what it means. An R-sq of 1 would mean that 100% of the variability in your dependent variable can be explained by your independent variable. With your R-sq of almost zero, it means Z is adding virtually nothing.

Thanks. That's what I thought, but the .05 increase of from X & Y to X, Y, & Z was throwing me off.

JTWood wrote:I know we have some stat heads here, and I need some help refreshing my memory on regression analysis. I have an output (N = 14) that I think might be based on three variables. When I ran the analysis, I got the following figures:

R-square = .86886 Intercept = -12.92890 Beta X = 13.1782 Beta Y = 0.80400 Beta Z = 16.5652

The Beta for Z seemed extraordinarily high to me, so I ran the analysis with just Z, and I got a number that seemed really odd to me:

R-square = 0.00363

So I ran the analysis again with X and Y, but without Z, and got:

R-square = 0.81803

Does this mean that variable Z is really doing nothing for my overall equation?

JTWood wrote:I know we have some stat heads here, and I need some help refreshing my memory on regression analysis. I have an output (N = 14) that I think might be based on three variables. When I ran the analysis, I got the following figures:

R-square = .86886 Intercept = -12.92890 Beta X = 13.1782 Beta Y = 0.80400 Beta Z = 16.5652

The Beta for Z seemed extraordinarily high to me, so I ran the analysis with just Z, and I got a number that seemed really odd to me:

R-square = 0.00363

So I ran the analysis again with X and Y, but without Z, and got:

R-square = 0.81803

Does this mean that variable Z is really doing nothing for my overall equation?

RyanK wrote:yeah.. thats some rough stuff.. you doing that for fun JT?

Doing it for our league. You saw that stuff that daullaz posted, right? We were trying to throw all of that into a formula to predict wins in a given season, or to at least determine what events were more important to determining wins. You know that managerial efficiency stat we posted? It's either worthless as a predictor, or I'm just not using it correctly.

I'm going to try some other stuff out. We'll see how it comes out...

WharfRat wrote:Did you guys go to school for that?

I have an Accounting degree, and I took a couple stat classes. The stuff in the first post is pretty basic. There's plenty more complicated stuff than what I posted. I just remember how to do this stuff because Excel will do all of calculations for you.