Dear all,
It is well documented that interaction terms in non-linear models might be hard to interpret.
Thanks toM.L. Buis (2010) "Stata tip 87: Interpretation of interactions in non-linear models", The Stata Journal, 10(2), pp. 305-308, I get that one should focus on odds ratio (multiplicative scale) to interpret interaction effects.
Issues are only raising when estimating marginal effects (additive scale). Thus the simplest way to make sure of the sign and significativity of the interaction effect is to run the model with the ",or" option to compute odds ratio. (As Marteen explained in a previous post for clogit, see link below)
However, such an option isn't available for a Nested Logit Model (nlogit , see help nlogit), while I guess the interaction interpretation issue remains.
How to compute odds ratio of nested logit model, or how to interpret properly interaction terms of a nested logit model?
Thanks a lot,
Charlie
It is well documented that interaction terms in non-linear models might be hard to interpret.
Thanks toM.L. Buis (2010) "Stata tip 87: Interpretation of interactions in non-linear models", The Stata Journal, 10(2), pp. 305-308, I get that one should focus on odds ratio (multiplicative scale) to interpret interaction effects.
Issues are only raising when estimating marginal effects (additive scale). Thus the simplest way to make sure of the sign and significativity of the interaction effect is to run the model with the ",or" option to compute odds ratio. (As Marteen explained in a previous post for clogit, see link below)
However, such an option isn't available for a Nested Logit Model (nlogit , see help nlogit), while I guess the interaction interpretation issue remains.
How to compute odds ratio of nested logit model, or how to interpret properly interaction terms of a nested logit model?
Thanks a lot,
Charlie
