Weighted fit in macro?

classic Classic list List threaded Threaded
3 messages Options
Reply | Threaded
Open this post in threaded view
|

Weighted fit in macro?

Petro
Hi list.
Is there any way to do weighted fit inside of the imageJ macro?
Thank you.
Petro

--
ImageJ mailing list: http://imagej.nih.gov/ij/list.html
Reply | Threaded
Open this post in threaded view
|

Re: Weighted fit in macro?

Michael Schmid-3
On 21/06/2018 14:41, Piter_ wrote:
 > Hi list.
 > Is there any way to do weighted fit inside of the imageJ macro?
 > Petro


Hi Petro,

sorry, there are currently no weights (or error bars) in the ImageJ
Curve Fitter. The only thing you can do is adding more reliable points
twice or three times.

For very few fit types where a "linear regression" method is available,
this method will result in different weighting than the one without
"linear regression".

E.g. "Exponential" gives equal weight (assumes equal errors) for all points.
"Exponential (linear regression)" is a least-squares fit to
   ln y = ln a + bx,
so it assumes the errors of ln y are equal. In other words, low y values
are considered more accurate (errors are assumed to be a fixed fraction
of the value).


Michael

--
ImageJ mailing list: http://imagej.nih.gov/ij/list.html
Reply | Threaded
Open this post in threaded view
|

Re: Weighted fit in macro?

Petro
Thanks Michael.
I will try it.
Best.
Petro

On Fri, Jun 22, 2018 at 11:36 AM, Michael Schmid <[hidden email]>
wrote:

> On 21/06/2018 14:41, Piter_ wrote:
> > Hi list.
> > Is there any way to do weighted fit inside of the imageJ macro?
> > Petro
>
>
> Hi Petro,
>
> sorry, there are currently no weights (or error bars) in the ImageJ Curve
> Fitter. The only thing you can do is adding more reliable points twice or
> three times.
>
> For very few fit types where a "linear regression" method is available,
> this method will result in different weighting than the one without "linear
> regression".
>
> E.g. "Exponential" gives equal weight (assumes equal errors) for all
> points.
> "Exponential (linear regression)" is a least-squares fit to
>   ln y = ln a + bx,
> so it assumes the errors of ln y are equal. In other words, low y values
> are considered more accurate (errors are assumed to be a fixed fraction of
> the value).
>
>
> Michael
>
> --
> ImageJ mailing list: http://imagej.nih.gov/ij/list.html
>

--
ImageJ mailing list: http://imagej.nih.gov/ij/list.html