On 13 Nov 2003 21:53:10 -0800, [EMAIL PROTECTED]
(Safa Gurcan) wrote:

> Dear Folks,
> In multiple regression analysis; model summary (R2, adjusted R2) and ANOVA 
> results display in the output screen. I used ridge regression analysis to 
> my data in SYSTAT but, didn't see any of the results above except 
> standardized and unstandardized coefficient and lamda . Does Ridge 
> regression have R2 and SEM ? How de we understand that the unstandardized 
> regression coefficients explained the variation of dependent variable?


Advice in  Tabachnick and Fidell (1996) is not enthusiastic
for ridge analysis.  Eventually, they recommend  using the 
computer package/ program,  BMDP-4R.   My old manual 
shows  a rather nice illustration of output.  Dated 1990, it is
possibly the newest manual, too.

(SPSS bought the BMDP company in 1991 or so; distribution
and development mainly stopped.  That is too bad, because
some BMDP  provided the best available output that I have
seen for several analyses:  repeated measures ANOVA;
correspondence analysis; loglinear models.)

This  program does show  *estimates*  of  R2 and of the 
residual SS (SEM)  as they vary with the ridge factor, and it
plots them, too.  The labels and descriptions in the manual
make it seem possible to me that those  R2 and SS  are 
not an *essential*  part of the ridge regression.  Or, to put it
another way, some folks might have another version, and that 
could be why your package does not provide them.
 - It does seem to me that you don't have much to point to, 
if you merely have a sequence of regression coefficients.

-- 
Rich Ulrich, [EMAIL PROTECTED]
http://www.pitt.edu/~wpilib/index.html
"Taxes are the price we pay for civilization." 
.
.
=================================================================
Instructions for joining and leaving this list, remarks about the
problem of INAPPROPRIATE MESSAGES, and archives are available at:
.                  http://jse.stat.ncsu.edu/                    .
=================================================================

Reply via email to