Failure to improve Likelihood after ... iteration

[ming-ming]

Hello,

I am estimating a non-linear system using full-information maximum likelihood. For some initial values, I got the warning "Failure to improve Likelihood after ... iteration". Could anybody explain to me what does that warning mean? Could the log likelihood be trusted in this case? I checked the user's guide and reference manual and there is no discussion on the system error information.

Many thanks~

Mingming

[trubador]

Some or all those initial values are probably far off from the optimum point and lead the estimation into regions where the likelihood function is poorly behaved. If you want to identify a solution that is either at a local or a global optimum, then you should find a way to ensure the convergence.