| Date: | Fri, 4 Jun 1999 13:14:30 -0600 |
| Reply-To: | "Raymond V. Liedka" <liedka@UNM.EDU> |
| Sender: | "SAS(r) Discussion" <SAS-L@UGA.CC.UGA.EDU> |
| From: | "Raymond V. Liedka" <liedka@UNM.EDU> |
| Subject: | Re: A regression question...... |
|
| In-Reply-To: | <19990604190144.19714.rocketmail@web308.yahoomail.com> |
| Content-Type: | TEXT/PLAIN; charset=US-ASCII |
|---|
On Fri, 4 Jun 1999, E. S. wrote:
> I have another question now, though, and it has to do
> with Proc REG: How do you force a regression analysis
> result to pass through a specific point???
> For example, say I want to run a regression of the
> form y=aX1 + bX2 + cX3, but I want to force the
> solution to pass through the point (x1, y1).
> Any ideas will be welcome, and thanks again,
> Ed
>
Ed,
In REG you can have the intercept estimated, or you can force the line
through the origin. You want neither...but force the line through (x1,
y1).
You can "trick" REG into this by transforming the data so that the
point (x1, y1) becomes the origin. In a DATA step, compute y*=y-y1 and
X1*=X1-x1. Now, if you tell REG the model is
Model y* = X1* X2 X3 / noint;
you will get the line forced through the point (x1, y1) and the regression
coefficients a, b, and c will not change. All you have done is shift the
location on two of the dimensions of your data (y, X1) in such a way that
(x1, y1) is the origin the X1-y plane.
ray
Raymond V. Liedka
Assistant Professor
Department of Sociology
University of New Mexico
|