Date: Tue, 20 Apr 2004 04:18:45 -0400
Reply-To: Subscribe Sas-L Anonymous <susi_i@HOTMAIL.COM>
Sender: "SAS(r) Discussion" <SAS-L@LISTSERV.UGA.EDU>
From: Subscribe Sas-L Anonymous <susi_i@HOTMAIL.COM>
Subject: Re: Orthogonal regression?
Hi,
Just read the posts here about orthogonal regression....
I need to perform this in order to obtain the slope for use in a device
calibration study.
It has been suggested that I use PROC CALLIS, but I don't really understand
the methodology. The PROC PRINCOMP code and output looks easier to
understand, however I don't know how to obtain the slope from this, and
don't have access to the book mentioned by Paige in the previous post.
Can anyone help?
Many thanks,
Susi
On Thu, 5 Jun 2003 08:34:04 -0400, Paige Miller <paige.miller@KODAK.COM>
wrote:
>simian wrote:
>> Hello all,
>>
>> I am trying to figure out how to do an orthogonal regression with two
>> variables in base SAS. The goal is to identify which sample points
>> fall outside of certain standard deviation bands (say, +/- 1 SD) as
>> measured by the perpendicular distance to a regression line. My
>> variables X and Y are in the same units.
>>
>> I understand that PROC ORTHOREG is not designed to do this. On my
>> data set, it produces essentially the same slope and intercept as PROC
>> REG.
>
>The orthogonal regression line can be determined from the first
>principal component, therefore you will need PROC PRINCOMP and some
>additional calcs that you will have to write yourself. The details can
>be found in Jackson, J. E. (1991) "A User's Guide to Principal
>Components", John Wiley and Sons, NY. See Chapter 15. Please note that
>formula 15.2.4 is incorrect due to typographical error.
>
>--
>Paige Miller
>paige.miller@kodak.com
>http://www.kodak.com
>
>"It's nothing until I call it!" -- Bill Klem, NL Umpire
>"When you get the choice to sit it out or dance, I hope you dance" --
>Lee Ann Womack
| 
