Statistics/Econometrics HELP PLEASE

[JimmyCracksCorn]

Originally posted by AnnC


It's not in Canada (if that's what you were thinking).

So where is it? I wasn't even aware you were Canadian.

[AnnC]

It's in a large US city.

(And I'm not Canadian, I'm a native of South Carolina).

[Al B. Sure!]

everything makes sense now.... got an R^2 value of .7667 and an SST of 255,094.288.

I then had to test H0: b2=b3=0 at the 5% significance level...

Fcritical= 3.32
F= [(255094.3-59513)/2]/(59513/30)= 97790.65/1983.76= 49.296 (did i make a mistake?)

Reject null.

and then I'm told the equation has been reestimated after adding squares and cubes of predictions and the new output is as follows:

R^2 0.594
S.E. of regression 38.538
Sum of square resid 37130
Mean Dep. Var 125.527 (same)
SD dep var 90.713 (same)
F-statistic 33.919 (BLAH?! I solved for the F-statistic by hand and got 8.4396)

F= [(59513(SSRrestricted)-37130)/2]/[37130/(33-5)= 8.4396

Fcritical=3.34

so reject the null... there are non-linear components.

something doesn't seem right about this though...

[AnnC]

Using your R^2 value of .7667 (which looks right to me, but I didn't verify the calculation), I got an F-statistic of 49.274 which is very close to your number. So the null is firmly rejected, and you can make a case that the independent variables do explain the y's.

In the second part, I'm not sure I understand what you mean by adding squares and cubes of predictions. Did your instructor give you a different model for that part, with more coeffcients to be estimated? Or is the second-part model a log-linear version of something like this?

y = B1 + B2*x2^2 + B3*x3^2 + e

[Al B. Sure!]

additional coefficients... the new model is:

y= b1+b2x2+b3x3+y^2+y^3

it is used to determine if additional, non-linear coefficients would be needed to better explain y.

[Kropotkin]

Originally posted by Albert Speer
so reject the null... there are non-linear components.

something doesn't seem right about this though...

No, it doesn't. Sorry if I **** it up even more, I'm a bit tired and am about to go to bed. Didn't take time to read everything tru. Just thought about looking up apolyton ot again.

I'm going with AnnC's model as your last seems a bit weird. What's the point of squared dependent variables as independent variables?

So you have a model with to independents with an Rsq of 0,77 and then add two more and get a Rsq of 0,6 or so (that shouldn't really happen, I guess it's the adjusted Rsq.). The reduced F-test implies that there's more info in the new independens, but the model loses a lot of power. You should really look at the partial t-tests and reduce the model again. If you know how to use t-tests, look at them. I prefer to look at 'em to using reduced F-tests.

[Odin]

AS is ALIVE!!!!!



We though you were dead.