EViews.com

I would like more control over the ETS modelling procedure. Specifically I would like to be able to explore how the value of the objective (eg AIC) responds to different starting points in the parameter space. In MS Excel I do this by randomly selecting starting values for the parameters and running several hundred iterations of the model search, and I find that the solution for a praticular time series is generally non-unique, unstable and sensitive to the starting point for the search over the parameter space. So for any given time series I would like to get a feel for the robustness of the EViews search procedure for an AIC-minimising (say) exponential smoothing model. While the EViews "ets" series proc (see page 495 of the Object Reference for EViews enables one to specify the value of parameters, and adopts the standard restrictions on the parameter space, it does not appear to allow exploration of the sensitivities described above. Thanks.

If I understand what you're asking for - you want to be able to randomize the starting parameters and then record the final objective values. You can do that in an EViews program easily enough.