ApolloChoiceModelling forum

Posted: **24 Oct 2020, 11:51**

Dear all,

is there a way to change the convergence criterium value (i.e., the value considered the gradient to be zero)? I want to demonstrate students that there might be differences in estimation between different software packages (often due to algorithmic choices and defaults). Example of mlogit() is attached.

Best
Sven

Posted: **09 Nov 2020, 18:04**

You can change the precision by changing the maxLik_settings, for example:

Code: Select all

apollo_estimate(apollo_beta, apollo_fixed, 
                apollo_probabilities, apollo_inputs, 
                estimate_settings = list(maxLik_settings = list(reltol=1.48e-08)))

Cheers
David

Posted: **09 Nov 2020, 18:21**

David,

Thank you so much!

Cheers
Sven

ApolloChoiceModelling forum

Adjust estimation (convergence) precision

Adjust estimation (convergence) precision

Re: Adjust estimation (convergence) precision

Re: Adjust estimation (convergence) precision