Page 1 of 1

Interpretation of the estimated parameters in MNL, MXL models

Posted: 05 Jul 2023, 15:47
by mateo.n
Dear Prof Hess and Dr Palma,

First of all, thank you for the great effort of making Apollo package and keeping it running together with this website. My question is very basic, so I am using the luxury of being anonymous online to ask.

Are the estimated parameters in MNL, MXL models log odds ratios by default? I was mostly working in STATA, and there is an emphasized difference between coefficients and odds ratios, in respect of using the relevant commands there...

Thank you and best regards,
Mateo

Re: Interpretation of the estimated parameters in MNL, MXL models

Posted: 05 Jul 2023, 16:14
by stephanehess
Mateo

the values you obtain are maximum likelihood estimates for model parameters. Their role depends on how you've specified your model. So in a simple linear in attributes MNL model, for example, the parameters would give you the unit impact of changes in the attribute on the utility

Stephane