MLE estimation in GAMS

This Max Likelihood Estimation example includes forming of confidence intervals. Standard errors are found by inverting the Hessian and normal quantiles are calculated using a minimal CNS model.

http://amsterdamoptimization.com/models/statistics/weibull.gms

This kind of analysis is only possible with new features in GAMS. However, it remains kind of klunky compared to Matlab, Gauss or SAS. The GAMS approach may be useful nevertheless, e.g. when using as part of a larger GAMS model (in which case calling a different system may be cumbersome), when the problem is very difficult (in which case stronger NLP solvers and automatic differentiation can help) or when you don't have access to these other systems.