Hi all ---
 
The power of learning to impose  restrictions on Regression/Linear Models is that you can create models that have the properties
that YOU WANT.   If YOU WANT TO CREATE A MODEL WITH "INTERESTING" PROPERTIES FOR YOU, THEN
DO IT.
 
It is quite simple to impose restrictions that force the least-squares function to pass  through any points that you like. 
Also, impose restrictions to force a polynomial to have any slope  at any point you would like, etc.
 
Any linear restrictions on the parameters are OK.
 
High School students can ENJOY using their algebra skills to impose their restirictions OF INTEREST or they can use SAS and allow the program to impose the restrictions.  Restrictions ("RESTRICT" in SAS) can be imposed to create ASSUMED MODELS that have desired properties and then further restrictions ("TEST" in SAS) can be imposed to TEST HYPOTHESES if desired.
 
Some simple examples of how this is done can be seen in INTRODUCTION  TO  LINEAR MODELS by Ward and Jennings.
 
--- Joe 
 
**********************************
Joe H. Ward, Jr.
167 East Arrowhead Dr.
San Antonio, TX 78228-2402
Phone: 210-433-6575
Fax:     210-433-2828
Email:  [EMAIL PROTECTED]
http://www.northside.isd.tenet.edu/healthww/biostatistics/wardindex
==============================
Health Careers High School
4646 Hamilton Wolfe Road
San Antonio, TX 78229
**********************************
 
----- Original Message -----
From: "Robert J. MacG. Dawson" <[EMAIL PROTECTED]>
To: "Anders" <[EMAIL PROTECTED]>
Cc: <[EMAIL PROTECTED]>
Sent: Thursday, November 20, 2003 8:00 AM
Subject: Re: Best straight line with one fixed point

>
>
> Anders wrote:
> >
> > I am using the least-mean-square method to calculate the best straight
> > line to a number of meassured points. But how do I change the formulas
> > if I want one of my meassured points to be fixed (the line must go
> > throu that point). This is possible in Excel, but only to fix the
> > y-value at x=0. I presume it is mathematically possible but havn't
> > found any formula for it.
>
> There are, as usual, infinitely many ways that you *can* do it. I can
> think of about three plausible ones off the top of my head.
>
> The question you need to answer first is - why do you want to do it?
> This affects what the correct way to do it would be (and whether it
> would be a smart thing to do at all!)
>
> -Robert Dawson
> .
> .
> =================================================================
> 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