Prev: Data acquisition tool kit
Next: Closing Excel doesn't work after saving
From: sunny on 17 May 2010 16:15
hi guys ,
probably didnt pose the right question.
i trying to generate a regression equation for response
(which is diurnal 24 hr concentration profile) and there are 5
(meteorological parameters) predictors with which i want to generate
the regression equation. out of which one predictor has some zero
value..

can i perform the operation to do the regression to
force the response to zero when this one parameter is also zero! (like
a logic algorithm) in Matlab
From: Tom Lane on 18 May 2010 11:42
> i trying to generate a regression equation for response
> (which is diurnal 24 hr concentration profile) and there are 5
> (meteorological parameters) predictors with which i want to generate
> the regression equation. out of which one predictor has some zero
> value..
>
> can i perform the operation to do the regression to
> force the response to zero when this one parameter is also zero! (like
> a logic algorithm) in Matlab

There are various ways you could accomplish this. Suppose you have
predictors X1 and X2, and you want the equation to be zero when X1 is zero.
Then the equation

Y = a*X1 + b*X1*X2 + error

has this property. Of course I have no idea whether this is a good model for
your data, but I hope this gives you something to think about. The basic
idea is that I don't include a constant term, nor an X2 term.

-- Tom


From: sunny on 19 May 2010 19:02
On May 18, 10:42 am, "Tom Lane" <tl...(a)mathworks.com> wrote:
> > i trying to generate a regression equation for response
> > (which is diurnal 24 hr concentration profile) and there are 5
> > (meteorological parameters) predictors with which i want to generate
> > the regression equation. out of which one predictor has some zero
> > value..
>
> > can i perform the operation to do  the regression to
> > force the response to zero when this one parameter is also zero! (like
> > a logic algorithm) in Matlab
>
> There are various ways you could accomplish this. Suppose you have
> predictors X1 and X2, and you want the equation to be zero when X1 is zero.
> Then the equation
>
>    Y = a*X1 + b*X1*X2 + error
>
> has this property. Of course I have no idea whether this is a good model for
> your data, but I hope this gives you something to think about. The basic
> idea is that I don't include a constant term, nor an X2 term.
>
> -- Tom

thanks tom.. that was helpfull .. i multiplied the other predictor
with the one that has zero value in it... if gave me a reasonable r.sq
value
thanks
 | 
Pages: 1
Prev: Data acquisition tool kit
Next: Closing Excel doesn't work after saving