Prev: xPC Real-time interrupt source
Next: Linear combination of linearly independent rows/columns of matrix
From: Konrad on 22 Apr 2010 07:37
Hi everyone,

I'm a MatLab greenhorn, so excuse me if my question seems too simple. I've been failing for hours to solve it.

Within a small MatLab routine, I'm fitting a linear function to a data set of four x and four y values, using coefficients = polyfit (x, y, 1). I'd also like to know the coefficient of determination, aka R-square, but I haven't found out how to get it without opening the cftool or plunging deep into the mysteries of structures and "fittype" etc. (I tried "f=fittype('a*x+b'), in order to then get the rsquare out of "fit", but it gives an error message.)

Thanks for help,
Konrad
From: Yi Cao on 22 Apr 2010 07:53
"Konrad " <Konrad.Lehmann(a)uni-jena.de> wrote in message <hqpcd1$8qu$1(a)fred.mathworks.com>...
> Hi everyone,
>
> I'm a MatLab greenhorn, so excuse me if my question seems too simple. I've been failing for hours to solve it.
>
> Within a small MatLab routine, I'm fitting a linear function to a data set of four x and four y values, using coefficients = polyfit (x, y, 1). I'd also like to know the coefficient of determination, aka R-square, but I haven't found out how to get it without opening the cftool or plunging deep into the mysteries of structures and "fittype" etc. (I tried "f=fittype('a*x+b'), in order to then get the rsquare out of "fit", but it gives an error message.)
>
> Thanks for help,
> Konrad

If you talk about the squared residual of the linear fitting, you can calculate it by

r = y - c(1)*x - c(2);
r2 = r.*r;

HTH
Yi
From: Konrad on 22 Apr 2010 08:24
Thanks, but your function returns the sum of squared errors. That is different from R² which nears 1 for a good fit.
Does R² have a similarly simple formula?

Thanks,
Konrad

"Yi Cao" <y.cao(a)cranfield.ac.uk> wrote in message <hqpdb1$oc4$1(a)fred.mathworks.com>...
> "Konrad " <Konrad.Lehmann(a)uni-jena.de> wrote in message <hqpcd1$8qu$1(a)fred.mathworks.com>...
> > Hi everyone,
> >
> > I'm a MatLab greenhorn, so excuse me if my question seems too simple. I've been failing for hours to solve it.
> >
> > Within a small MatLab routine, I'm fitting a linear function to a data set of four x and four y values, using coefficients = polyfit (x, y, 1). I'd also like to know the coefficient of determination, aka R-square, but I haven't found out how to get it without opening the cftool or plunging deep into the mysteries of structures and "fittype" etc. (I tried "f=fittype('a*x+b'), in order to then get the rsquare out of "fit", but it gives an error message.)
> >
> > Thanks for help,
> > Konrad
>
> If you talk about the squared residual of the linear fitting, you can calculate it by
>
> r = y - c(1)*x - c(2);
> r2 = r.*r;
>
> HTH
> Yi
From: John D'Errico on 22 Apr 2010 08:24
"Konrad " <Konrad.Lehmann(a)uni-jena.de> wrote in message <hqpcd1$8qu$1(a)fred.mathworks.com>...
> Hi everyone,
>
> I'm a MatLab greenhorn, so excuse me if my question seems too simple. I've been failing for hours to solve it.
>
> Within a small MatLab routine, I'm fitting a linear function to a data set of four x and four y values, using coefficients = polyfit (x, y, 1). I'd also like to know the coefficient of determination, aka R-square, but I haven't found out how to get it without opening the cftool or plunging deep into the mysteries of structures and "fittype" etc. (I tried "f=fittype('a*x+b'), in order to then get the rsquare out of "fit", but it gives an error message.)
>
> Thanks for help,
> Konrad

I'm pretty sure that I return the R^2 parameter from
polyfitn. So you could use that tool instead of
polyfit.

http://www.mathworks.com/matlabcentral/fileexchange/10065

HTH,
John
From: Konrad on 22 Apr 2010 09:03
Hi John,

I've downloaded your function, but don't see how I get R² out of it. Could you please indicate how it is done?

Best wishes,
Konrad
 |  Next  |  Last
Pages: 1 2
Prev: xPC Real-time interrupt source
Next: Linear combination of linearly independent rows/columns of matrix