The confidence intervals, fminsearch

[Alexey]

Dear all,

I am fitting the experimental data to non-linear fucntion. The only
algorithm that I found is working well is implemented in fminsearch.

I would like to estimate the confidence intervals for the parameters
of the model
however fminsearch does not output either Jacobian matrix or
residuals
Any suggestions on how I can use fminsearch and estimaet confidence
intervals?

Thank you !

PS Yes I found that NLPARCI directly calculating Confidence intervals
on parameters of nonlinear models. CI = NLPARCI(BETA,RESID,J) returns
the 95% confidence interval CI. However NLINFIT does not fit my data
so I can not really use NLPARCI

[Mary Ann Freeman]

In article <eedfe...@webx.raydaftYaTP>, a_va...@yahoo.com says...

You could try calling FMINSEARCH to get your solution, then pass that
solution as a start point, the 4th argument beta0, to NLINFIT to get a
Jacobian, then call NLPARCI with the Jacobian. (Of course, the function
you pass to FMINSEARCH is the square of what you pass to NLINFIT, but if
you already tried NLINFIT you probably have this part figured out.)

HTH,
-- Mary Ann