Hi AlL, I ahve this problem that my objective function is discontinous in the paramaters and I need to use methods such as nelder-mead to get around this. My question is: How do i compute standard errors to a problem that does not have a gradient? Any literature on this is greatly appreciated. Jean,