choice-metrics.com

Hello,

With 3000-row data I created externally to run the explicit partial profiles design, I get the MNL D-Error as 7374. How can we check against this value to understand whether our design is good enough?

Thank you very much!

Beyza

Such a large D-error is very suspicious and indicates a very poorly identified model. It could be that you specified utility functions unidentified parameters, or you could have included correlated attribute levels in your candidate set. A D-error is usually below 1 or 2, anything above it is not a good sign.

Michiel