Stata The Stata listserver
[Date Prev][Date Next][Thread Prev][Thread Next][Date index][Thread index]

RE: st: RE: problem with ridge regression output


From   "FEIVESON, ALAN H. (AL) (JSC-SK) (NASA)" <[email protected]>
To   "'[email protected]'" <[email protected]>
Subject   RE: st: RE: problem with ridge regression output
Date   Wed, 8 Dec 2004 08:55:55 -0600

Marcello - Yes - I forgot to mention that the added "Y's" need to be zero.
With respect to your second point, if you add (c^2)*I to the X'X matrix is
the ridge parameter defined as c or c^2? I don't remember the exact
definition. If it's c, then you are right - the identity should be
multiplied by sqrt(c). Thanks for your clarification.

Al



-----Original Message-----
From: [email protected]
[mailto:[email protected]]On Behalf Of Marcello
Pagano
Sent: Wednesday, December 08, 2004 8:49 AM
To: [email protected]
Subject: Re: st: RE: problem with ridge regression output


First, it should be stated that the added observations are zeroes (the 
ys=0).
Second, is it the ridge parameter or its square root that you want?

m.p.

FEIVESON, ALAN H. (AL) (JSC-SK) (NASA) wrote:

>Jade - I agree with you. Ridge regression is supposed to effectively add a
>positive constant to the diagonal of the cross-product matrix, so even if
>that matrix were non-full rank to begin with, it should be full rank after
>adding the ridge parameter. Of course you could always do ridge regression
>"by hand" by appending p extra observations consisting of the ridge
>parameter times an identity matrix, where p is the number of variables in
>the regression, and then doing OLS.
>
>  
>
*
*   For searches and help try:
*   http://www.stata.com/support/faqs/res/findit.html
*   http://www.stata.com/support/statalist/faq
*   http://www.ats.ucla.edu/stat/stata/
*
*   For searches and help try:
*   http://www.stata.com/support/faqs/res/findit.html
*   http://www.stata.com/support/statalist/faq
*   http://www.ats.ucla.edu/stat/stata/



© Copyright 1996–2024 StataCorp LLC   |   Terms of use   |   Privacy   |   Contact us   |   What's new   |   Site index