On 22 Feb 2004 19:23:44 -0800, [EMAIL PROTECTED] (chokie) wrote:

> I have a qns, let say an original Regesssion to is be estimated as
> follows:
> 
> Y = B + b1x1 + b2x2
> 
> Then, if diagnostic test suggests that the relationship between Y and
> x1 is nonlinear, add a variable X3 (which is square of X1).
> 
> Y = B + b1x1 + b2x2 + b3(x1*x1)
> 
> 
> Qns: how to interpret X1 since it has both a linear and nonlinear
> relationship
 [ snip, rest]

Was the non-linear shape evident from plotting the raw data?

The fitted regression equation gives you a new, theoretical set of
points to plot.  Before you accept the 'statistical improvement' as
truly being an improvement, you should make sure that the
equation still makes sense.  

 - I read a published example once that was totally incompetent,
and it was intended to *illustrate*  using a squared component!
 - the 'fit'  line passed through several points near zero, 
arched high where there were no points at all, and descended 
steeply to pass through one high x-value.  The residuals were
pretty small.  The fitted line had no useful meaning, except 
that it was a 'line with small residuals'.

So, make sure the fit makes sense for your question, if
you want more than 'small residuals'.

-- 
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