Hello Chenge Firstly, thank you for your interest and useful helping. i use the following function. i use datas to evaluate the function a=[636 720 768 688 712 624 772 748 632 676 704 760 668 700 728 652 692 636 596 808 ] %%population data for j=1:(length(a)-1) f=@(x)(((a(j+1)-x(1)*a(j)*exp(a(j)/(x(2)))+a(j))^2)/(2*x(3)*x(3)*x(1)*a(j)*exp(-a(j)/(x(2))))+log(sqrt(2*3.141592654*x(3)*x(3)*x(1)*a(j)*exp(-a(j)/(x(2))))));
i am using " flog " function for fminsearch. this is the log likelihood function getting from the following population model; a(t+1)=f(a(t))+s*sqrt(f(a(t)))*epsilon(t) , epsilon(t)~N(0,1), f(a(t))=b*n*exp(a(t)/n)
so i want to estimate b,n,s parameters, Namely respectively x(1),x(2),x(3)
also thank you for the paper, but i can't access for free, are there another way to dowload it for free?
> > Hello, > > What function are you currently using fminsearch on? I might try something like this: > > function estimator = estimator(x, data) > if x outside valid range, estimator = infinity > else estimator = regular MLE estimator > > (By the way, this is the MLE estimator I use and it's always worked well for me: http://www.nature.com/nmeth/journal/v7/n5/full/nmeth0510-338.html ) > > As for the other part of your question, I'm not quite sure what you're asking, but you can set the termination tolerance of your parameters with TolX in fminsearch's options. > > Hope this helps somewhat! > > Chenge