Thread Subject: Nonlinear Least Squares Data fit

I am using a Nonlinear Leastsqare fit algorithm from the matlab library to fit the curves in my data using a Maxwellian distribution.
Fit equation ~
 d1*(m/T1)^1.5*(exp(.5m(v-u)^2)/T1))+d2*(m/(T2)^1.5*(exp(.5m(v-u)^2/T2))

There are 5 constrain fit parameters in the equation (d1, d2, T1, T2, ul).
The problem I am having is that there doesn't seem to be one specific set of constraints that work for the model. Each time a run the program I can get some wildly different values for my constraints. I have a pretty good idea of the approximate values of each parameter. I was wondering if anyone had some good ideas on either a better fitting algorithm, or suggestions on how to better run on constrain the nonlinear least squares fit algorithm?

What matlab function do you use? do you specify the starting
values for your parameters?
You can try the Ezyfit toolbox:
http://www.mathworks.com/matlabcentral/fileexchange/10176


"Meredith " <kepera@gmail.com> wrote in message <h6pvvp$prk$1@fred.mathworks.com>...
> I am using a Nonlinear Leastsqare fit algorithm from the matlab library to fit the curves in my data using a Maxwellian distribution.
> Fit equation ~
> d1*(m/T1)^1.5*(exp(.5m(v-u)^2)/T1))+d2*(m/(T2)^1.5*(exp(.5m(v-u)^2/T2))
>
> There are 5 constrain fit parameters in the equation (d1, d2, T1, T2, ul).
> The problem I am having is that there doesn't seem to be one specific set of constraints that work for the model. Each time a run the program I can get some wildly different values for my constraints. I have a pretty good idea of the approximate values of each parameter. I was wondering if anyone had some good ideas on either a better fitting algorithm, or suggestions on how to better run on constrain the nonlinear least squares fit algorithm?