Mallows Cp* statistic
Magenta
cheers
Neil W. Henry
Magenta wrote:
> How do I get this statistic from linear regression in SPSS version 10?
>
> cheers
As far as I can tell Cp is not reported by SPSS. Of course SPSS regression
has no routine for "all subsets regression", so Cp is not relevant to the
kind of use SPSS assumes its clients will have for the regressions. Since
Cp is easily derived from an F statistic, or from Residual Sums of
squares, it is not a problem to compute one of them. (i.e. the C for a
specific subset of a "full" set of predictors).Of course if you want to
get Cp for everyone of the zillion or so subsets of a set of several
hundred predictors, that's not a happy alternative.
See p. 190 of Weisberg "Applied Linear Regression" 1980 , or more
generally, all of his Chapter 8.
--
*************************************************
`o^o' * Neil W. Henry (nhe...@vcu.edu) *
-<:>- * Virginia Commonwealth University
*
_/ \_ * Richmond VA 23284-2014 *
*
*************************************************
Magenta
statistic would be a good feature to have in SPSS. Minitab does it.
cheers
Michelle
"Neil W. Henry" <nhe...@vcu.edu> wrote in message
news:3BA8FEAD...@vcu.edu...
Rich Ulrich
<mag...@paradise.net.nz.nospam> wrote:
> SPSS developers please take note: All subsets regression with Mallow's Cp
> statistic would be a good feature to have in SPSS. Minitab does it.
>
All subsets regression is (at least generally) a mistake of amateurs.
Do you want Cp for any other purpose?
I hope that the results of "all subsets regression" should
become increasingly hard to get published, since the time
of its introduction and minor faddishness in the 1970s.
--
Rich Ulrich, wpi...@pitt.edu
http://www.pitt.edu/~wpilib/index.html
Magenta
"Rich Ulrich" <wpi...@pitt.edu> wrote in message
news:k32krt8d8d52as2ka...@4ax.com...
> On Sat, 22 Sep 2001 07:21:27 +1200, "Magenta"
> <mag...@paradise.net.nz.nospam> wrote:
>
> > SPSS developers please take note: All subsets regression with Mallow's
Cp
> > statistic would be a good feature to have in SPSS. Minitab does it.
> >
>
> All subsets regression is (at least generally) a mistake of amateurs.
> Do you want Cp for any other purpose?
Yep, I want Mallow's Cp for its help in assessing the value of a multiple
regression.
> I hope that the results of "all subsets regression" should
> become increasingly hard to get published, since the time
> of its introduction and minor faddishness in the 1970s.
The Cp option seems tied to the all subsets regression request in other
packages (e.g. Minitab). And I have *no* problem with all subsets
regression when one is working with an extremely complex regression - give
us all the tools that make our jobs easier when moving from theory to
statistical testing.
I was also wondering about asking for the PRESS statistic - but I haven't
read up on that one lately.
cheers
Michelle
Raynald Levesque
>> <mag...@paradise.net.nz.nospam> wrote:
>>
>> > SPSS developers please take note: All subsets regression with Mallow's
>Cp
>> > statistic would be a good feature to have in SPSS. Minitab does it.
>> >
>>
Hi Michelle
You might be interested to look at the file
"Do All-Subset regressions.SPS" which you will find in
http://pages.infinit.net/rlevesqu/SampleSyntax.htm#RegressionRepeatedMeasures
Mallows CP statistics is NOT computed however if you provide me with
the formulas to compute it, I will add that project on my list of
things to do.
Raynald Levesque rlev...@videotron.ca
Visit My SPSS Pages: http://pages.infinit.net/rlevesqu/index.htm
celalg...@gmail.com
> How do I get this statistic from linear regression in SPSS version 10?
>
> cheers
If you follow "Regression > Linear > Paste" you will see the syntax window. Then, type SELECTION end of your /STATISTICS line.
In addition to CP, it also provides 3 more selection criteria.
But, the problem is that SPSS and Minitab provide different Cp values !!!
Good luck !!