how to set nlinfit parameters ?J

Hello

I would need some help for fitting my data. Let's assume that I want to fit data registred in the matrix "spe":

x=spe(N,1)
y=spe(N,2)

and that it should be fitted by a Gaussian function

I have defined the matlab function for the Gaussian :

function fun = gauss(b1,b2,b3,x)
    fun = b3 * (1/(b1*sqrt(2*pi))) * exp(-(x-b2).^2/(2*(b1).^2));
end

and then, I want to use this function to fit my spe data using this expression in my main program using the nlinfit function.

As far as I understood, I have to create a function handle, which, by the way I am not understanding exactly what it is, in order to use nlinfit :

I tried :

mdl = @gauss
b0 = [1 1480 500];
b = nlinfit(spe(:,1), spe(:,2), mdlgausstest,b0);

but it is returning some error message.

Could you help me? Thank you very much !!

Answer 1

[Star Strider]

I’m in a context here that is very much less than clear.

The Gaussian function is defined as:

Gauss = @(a,b,c,x) a*exp(-((x-b)/(2*c)).^2);

or in terms that the nlinfit function will understand:

Gfcn = @(p,x) p(1).*exp(-((x-p(2))./(2*p(3))).^2);

Use the ‘Gfcn’ function as your objective function with nlinfit and all should be well.

Your ‘mdl’ function makes absolutely no sense to me, so I won’t essay to correct it.

So just use ‘Gfcn’, and — with the correct initial parameter estimates vector — you should have no problems fitting it to your data, if they are truly described by a Gaussian function.

Note — I do not have your data to test this with, so I am listing it as UNTESTED CODE. It should work.

Answer 2

[Anna Levy]

Thanks for your answer. I was trying to find a way to use nlinfit in a different way. But anyway, I found out my solution.

Sorry my question was really not clear.

[Star Strider]

My pleasure.