[Scilab-users] Scilab leastsq exponential fitting

[Antoine Monmayrant]

On 07/28/2014 03:59 PM, chloe.kykam wrote:
> Thanks Antoine, it does work after I restart Scilab.
>
> However, my actual code consists of more than 1 data set, and I would have
> to normalise them according the their respective areas under the curve. The
> code only works without the normalisation part (replacing ri with r), but
> error 21 arises when the part is included.
>
> funcprot(0)
> workbook = readxls('C:\Users\Kying\Desktop\Stray_light\Data.xls')
> sheet = workbook(1) //old
>
> k=6.63e-34*3e8/1.38e-23
> x=sheet(:,2)
> ri=sheet(:,3)
> gi=sheet(:,4)
> bi=sheet(:,5)
> uvi=sheet(:,6)
>
> //normalization
> norm=max([inttrap(x,ri),inttrap(x,gi),inttrap(x,bi),inttrap(x,uvi)])
> r=norm/inttrap(x,ri)*ri
> g=norm/inttrap(x,gi)*gi
> b=norm/inttrap(x,bi)*bi
> uv=norm/inttrap(x,uvi)*uvi
> //end of normalization
>
> dims=1
> ur=min(find(r>0.71*max(r))) //uppler limit
> lr=max(find(find(r<10)<find(r==max(r)))) //lower limit
> wr=matrix((cat(dims,zeros(lr-1,1),ones(ur-lr+1,1),zeros(1024-ur,1))),1024,1)
> //weighting
>
> //exponential function
> function y=yth(x,a)
>       y=a(1)*exp(-k./x/a(2))
> endfunction
>
> //initial parameters
> a0=[1.0;300]
>
> function e=myfun(a,x,r,wr)
>       e=wr.*(yth(x,a)-r)
> endfunction
>
> //call leastsq
> [f,xopt,gopt]=leastsq(list(myfun,x,r,wr),a0)
>
> scf();
> plot2d(x,r,logflag="nl",style=[color("red")])
> //plot2d(x,yth(x,xopt))
> plot2d(x,g,logflag="nl",style=[color("green")])
> plot2d(x,b,logflag="nl",style=[color("blue")])
> plot2d(x,uv,logflag="nl",style=[color("purple")])
>
> Error message:
> -->exec('C:\Users\Kying\Desktop\Stray_light\testing2.sce', -1)
>   !--error 21
> Invalid index.
> at line     118 of function numderivative called by :
> at line       2 of function Dfun called by :
> at line       2 of function %opt called by :
> at line      92 of function leastsq called by :
> [f,xopt,gopt]=leastsq(list(myfun,x,r,wr),a0)
> at line      38 of exec file called by :
> exec('C:\Users\Kying\Desktop\Stray_light\testing2.sce', -1)
>
> Do you have any idea why that is?

Check the length and dimension (ie is it 1xn or nx1) of the vectors you 
are using with and without the normalization.
Are there any difference before and after the normalization?
leastsq is particularly picky: x and y need to be column vectors, not 
row vectors.
Moreover, when x and y have not the right length or dimensions, it tends 
to output cryptic error messages.

Cheers,

Antoine

>
> Thanks
>
>
>
>
> --
> View this message in context: http://mailinglists.scilab.org/Scilab-leastsq-exponential-fitting-tp4030949p4030973.html
> Sent from the Scilab users - Mailing Lists Archives mailing list archive at Nabble.com.
> _______________________________________________
> users mailing list
> users at lists.scilab.org
> http://lists.scilab.org/mailman/listinfo/users
>