ApolloChoiceModelling forum

Hi Prof. Hess,

In example 16, for the business shift for the random coefficient "v_tt", you use multiplication instead of addition, would you share your consideration for such a specification?

randcoeff[["v_tt"]] = ( exp( mu_log_v_tt
+ sigma_log_v_tt_inter * draws_tt_inter
+ sigma_log_v_tt_inter_2 * draws_tt_inter ^ 2
+ sigma_log_v_tt_intra * draws_tt_intra )
* ( gamma_vtt_business * business + ( 1 - business ) ) )

or

randcoeff[["v_tt"]] = ( exp( mu_log_v_tt
+ sigma_log_v_tt_inter * draws_tt_inter
+ sigma_log_v_tt_inter_2 * draws_tt_inter ^ 2
+ sigma_log_v_tt_intra * draws_tt_intra )
+ ( gamma_vtt_business * business ) )

Best,
Wenjian

Hi

using multiplication in this way implies that you're multiplying the entire distribution, rather than just shifting the mean of the log of the VTT. But you can easily compare the two and see which gives more reasonable results. With the additive one, you'd need to consider whether you want the shift inside or outside the exp

Stephane