Relationship between HB and MNL in ACBC

Question

Is there any relationship between HB and MNL in ACBC? If using HB to estimate individual partworths, does ACBC still use the MNL model  to influence the estimated partworth utilities (apart from looking for interaction effects)?

Thanks for your support and kind regards, Tim

Answer 1

Tim,

Both the HB and the aggregate MNL run from the same data, so there is that relationship.  But each model is run independently and neither uses the other as an input.

Comments

Thank you for that. What do we need MNL for then? HB is estimating individual partworth utilities... where exactly during the process of estimating anything does ACBC use the aggregate MNL?

---

I think the only place ACBC uses aggregate MNL is in the interaction search tool - aggregate MNL has a stat test, the -2 log likelihood test, that's useful for testing whether interactions are significant.  Aggregate MNL is not one of the estimation methods that will give you a set of utilities.  

Of course "HB" is just shorthand for "HB-MNL" and HB-MNL is a member of the MNL family of models.  Similarly in analyzing CBC experiments you can choose to run latent class, but that's a latent class MNL model.  I hope this helps.