Thanks for the replies. It was the difference between the adjusted R2
calculations that I was concerened about. I have also tried the
calculations in SAS JMP and that does not output R2 or adj R2 for
regressions without a constant term. Maybe this is a more appropriate
thing to do as it avoids the confusion with interpretation you guys
have brought to my attention.
Many thanks,
Dr L Green.
In article <[EMAIL PROTECTED]>,
[EMAIL PROTECTED] (Daniel J. Nordlund) wrote:
> Rich Ulrich wrote:
> >On Thu, 24 Feb 2000 12:45:29 GMT, [EMAIL PROTECTED] wrote:
> >
> >> I have found a difference between the results produced by SPSS and
> >> SYSTAT in linear regression with no constant term. Below are the
> >> results from the programs. As you can see the adjusted R2 given
by the
> >> 2 programs is different. Which one is correct?
> > < snip, data >
> >> Linear regression Y on X with no constant term.
> >>
> >> SYSTAT:
> >> R2 = 0.999365492
> >> adj R2 = 0.999365492
> >>
> >> SPSS:
> >> R2 = 0.9993654922987
> >> adj R2 = 0.9993020415285
> >
> >You may note:
> >
> > a) SYSTAT shows the same number twice: that is, there is no
> >suggestion of "adjustment".
> > b) This is regression "with no constant term." Okay, that is the
> >*abnormal* way to do regression, which sacrifices (for one thing)
the
> >ease of referring to R-squared. Are you talking about accounting for
> >the SS around the MEAN, or around ZERO?
> > c) I prefer to continue talking about SS around the MEAN, and if you
> >did that, then you *might* continue ( - but, no assurances) to use
the
> >ordinary 'shrinkage' formula to get an "adj R2".
> > d) When you have those values of .999+, as above, the easy way to
get
> >them (with all-positive numbers) is to have the SS around ZERO. I
> >*assume* that the usual shrinkage formula has to be modified, if
> >anyone has come up with a formula at all.
> >
> >Thus, I conclude that SYSTAT decided not to provide an adjusted R2,
> >possibly because there is nothing reliable or conventional in the
> >literature.
> >
> >SPSS probably provided one, if their distinction is not round-off
> >error. I would want to check their references before I trusted in
> >reporting it; but I am concerned, that almost no one uses that
> >computation (through the origin/ SS around zero), and that was not
> >emphasized in the question. If using that if it is not a
convention,
> >already, in your area of specialty, then you have enough troubles
> >explaining why you used that regression ... without getting into the
> >details of *that* adjusted R-squared.
> >
> >--
> >Rich Ulrich, [EMAIL PROTECTED]
> >http://www.pitt.edu/~wpilib/index.html
> >
>
> The adjusted R-squared value reported by SPSS is
>
> R2adj = 1-(MSres/MStot)
>
> for the regression model without an intercept. Whether the adjusted
R-squared
> is useful or not in the no intercept model is an open question.
>
> Dan Nordlund
>
Sent via Deja.com http://www.deja.com/
Before you buy.
===========================================================================
This list is open to everyone. Occasionally, less thoughtful
people send inappropriate messages. Please DO NOT COMPLAIN TO
THE POSTMASTER about these messages because the postmaster has no
way of controlling them, and excessive complaints will result in
termination of the list.
For information about this list, including information about the
problem of inappropriate messages and information about how to
unsubscribe, please see the web page at
http://jse.stat.ncsu.edu/
===========================================================================
