| linfit and regress questions [message #56693] |
Sun, 11 November 2007 08:27  |
rlayberry
Messages: 33
Registered: November 2004
|
Member |
|
|
Hi
I am interested in fitting a line of the form y=mx+c using regress or
linfit. I want the option of fixing the value of c at zero. Can I do
this in either of these 2 functions or would I need to use something
else?
Thanks
Russ
|
|
|
|
| Re: linfit and regress questions [message #56766 is a reply to message #56693] |
Tue, 13 November 2007 08:14  |
Craig Markwardt
Messages: 1869
Registered: November 1996
|
Senior Member |
|
|
rlayberry@hotmail.com writes:
> On 13 Nov, 03:16, Craig Markwardt
> <craigm...@REMOVEcow.physics.wisc.edu> wrote:
>> hradilv <hrad...@yahoo.com> writes:
>>
>>> Transform the data?
>>
>>> mhat = total(x*(y-c1))/total(x*x)
>>
>> Of course this all assumes the data are meant to be equally weighted
>> (no error bars, or all error bars are equal).
>>
>
> and does anyone know how the chi-sq or any other goodness of fit
> paramater would be calculated?
Do you mean, how to compute the chi-square value? Probably something
like,
TOTAL( ((Y-MHAT*X)/ERROR)^2 )
But you've never really specified what you are trying to do. Why not
just use a fitting program like MPFITFUN or CURVEFIT and specify
whatever model and/or error bars that you want?
Craig
|
|
|
|
| Re: linfit and regress questions [message #56768 is a reply to message #56693] |
Tue, 13 November 2007 02:24 |
rlayberry
Messages: 33
Registered: November 2004
|
Member |
|
|
On 13 Nov, 03:16, Craig Markwardt
<craigm...@REMOVEcow.physics.wisc.edu> wrote:
> hradilv <hrad...@yahoo.com> writes:
>
>> Transform the data?
>
>> mhat = total(x*(y-c1))/total(x*x)
>
> Of course this all assumes the data are meant to be equally weighted
> (no error bars, or all error bars are equal).
>
and does anyone know how the chi-sq or any other goodness of fit
paramater would be calculated?
russ
|
|
|
|
comp.lang.idl-pvwave archive
Members
Search
Help
Login
Home
