Issues using fsolve while iterating

Hi, I'd appreciate any suggestion(s) on the following:

I'm using mle for my estimation but before evaluating my likelihood function I have to solve for an auxiliary parameter, r, that can be solved from a model condition given theta (theta1 and theta2).

Given theta1 and theta2 I have to solve for r from the following equation:

y = (((pmin-r)*100)/(((pmax - r)^2)*gampdf(gaminv((pmax-pmin)/(pmax-r),theta1,theta2),theta1,theta2)))-1;

In every iteration random numbers will be generated for both theta1 and theta2 within the specified bounds in fmin. I have to modify the code to take into account the following two issues:

1.Given theta1 and theta2 the equation above might not have a solution.

2.For fsolve’s initial guess the function might not be defined. It would help to restrict r to be within (0, pmin) but I can’t specify bounds for fsolve.

thanks in advance,

Jose

Answer 1

[Matt J]

Use LSQNONLIN instead, which does let you specify bounds. Use the exitflag and the resnorm output arguments to determine if a solution was found.

Answer 2

[Jose Pellerano]

thanks Matt, it seems to work.

Jose

[José-Luis]

Please accept an answer if it helped you.