Comparing simulated annealing results w/default optimization

[tds]

I have been using simulated annealing when a model doesn't converge properly using the default optimization method. Then I compare the results of the models run with simulated annealing to the results of the other models run with the default optimization method.

Is there any reason why it would be necessary to use simulated annealing on all models in the data set, or is OK to compare models in the data set optimized with the two different methods?

Thank you.

[cooch]

I have been using simulated annealing when a model doesn't converge properly using the default optimization method. Then I compare the results of the models run with simulated annealing to the results of the other models run with the default optimization method.

Is there any reason why it would be necessary to use simulated annealing on all models in the data set, or is OK to compare models in the data set optimized with the two different methods?

Thank you.

You answer your own question in the first sentence. If any particular model shows convergence issues (or where you have reason to suspect there might but multiple modes in the likelihood), try SA. Thats it...