ApolloChoiceModelling forum

Dear Concern,

With reference to the example of the Hybrid choice model with Ordered Logit, I have two questions.

1. While choosing to do the Hybrid choice model in order to understand the respondent's latent behaviors through attitudinal questions in a SP surey, I did not understand the example of drug choice data, where the estimation gives the result of latent factors (the attitudinal factors, for example: zeta_quality>0.933171 and subsequently the followings:
tau_quality_1 -1.917914
tau_quality_2 -1.016011
tau_quality_3 1.119695
tau_quality_4 2.246541)

I believe there is a starting value which we define prior to estimation. But what should be the starting value and why does the value fluctuates between +- 1& 2 ?
tau_quality_1 =-2, tau_quality_2 =-1, tau_quality_3 = 1, tau_quality_4 = 2.. ?

2. While I understand using SEM (structural equation modeling) which enables to unveil the latent constructs as well, does the SEM fit and HCM model by apollo package setting estimates gives the different values for the factors incorporated in estimation ?

Any comment for clarification is highly appreciable !

Hi

on point 1, these are the starting values for the thresholds for the ordered logit model, which by definition need to be monotonically increasing.

Related to this and also point 2, the zeta parameters are not the attitudinal factors, but the impact of the latent variable in the measurement model for the attitudinal statements.

The actual latent variables are of course latent/unobserved and random, but you can look at the posterior distributions for them using apollo_conditionals after estimation

Stephane