Dear Sankara
I would also advise you to do a residual analysis
regardless what value of n maximize R. You want to
make shure that the basic assumptions of the model
have been satisfied so you can have enough confidence
to use your model for prediction.
Dr. Jaime Curts
University of Texas Pan American
Edinburg, Texas. 78539
USA
-- Sankara narayanan <[log in to unmask]> wrote:
> Hi Micheal,
>
> Thank you for your reply.
>
> I tried the solution by using the same method that
> you have suggested.
> I assumed a value of n so that Z = X^n. Now it forms
> an ordinary linear
> equation Y = A + B*Z. I tried by assuming various
> values of n. I plotted a
> graph with n values on the x-axis and R square
> values on the y-axis.
> Finally I chose the value of n that gives maximum R
> square.
>
> With that value of n, I was able to estimate A and
> B. The value of n I chose
> is 0.196
>
> I was able to do this as I had an idea about the
> range of n. But if I don't
> have an inkling about the range of n, such
> assumptions may make my
> predictions erroneous.
>
> To make sure that my predictions are correct, can
> you tell me the parameters
> that should be investigated (apart from R square) ?
>
> Some say that one should not go blindly by the value
> of R square. The
> difference in predicted and actual values should be
> considered.i.e, Standard
> Error should be considered.
>
> Can you suggest some tests to make sure that
> predicted n value is correct ?
>
> Thanks, Sankar.
>
>
>
>
>
> >From: "Graziano, Michael" <[log in to unmask]>
> >Reply-To: Concerned with the initial learning and
> teaching of statistics
> > <[log in to unmask]>
> >To: [log in to unmask]
> >Subject: Re: Polynomial Regression
> >Date: Mon, 19 Mar 2001 13:27:11 -0500
> >
> >Hi Sankara,
>
> You can transform the data as follows. For each x,
> define z = x^n. Then,
> if you work with the z's you have Y = A + B*Z which
> is solved via OLE
> (Ordinary Linear Estimation). The A and B estimated
> in this
> >fashion are also the A and B required in the
> original problem.
>
> >Hope that helps.
> >
> >Rgds
> >
> >mjg ([log in to unmask] ; 203-353-8100 x277)
> >
>
>=======================================================================
> >
> >I have a set of values for Y and X. Y is the
> response variable and X is the
> >independent variable.
> >
> >The trend is as follows:
> >
> >Y = A + B*X^n
> >
> >I want to get the values of A,B and n. Can you help
> me in obtaining the
> >regression solution ?
>
>
_________________________________________________________________________
> Get Your Private, Free E-mail from MSN Hotmail at
http://www.hotmail.com.
__________________________________________________
Do You Yahoo!?
Get email at your own domain with Yahoo! Mail.
http://personal.mail.yahoo.com/
|