ApolloChoiceModelling forum

I was looking at the hybrid choice modelling example from the following paper.

Hess, S., & Hensher, D. A. (2013). Making use of respondent reported processing information to understand attribute importance: a latent variable scaling approach. Transportation, 40(2), 397-412.
https://link.springer.com/article/10.10 ... 012-9420-y

I have one or two questions regrading the estimation.

1) In the paper we can see that the hybrid model is developed by integrating the probabilities of equation (7) and equation (8), can we first do the
analysis from equation (7) separately and the analysis for the equation 8) separately?

2) Is there a example of such model in combination ? Its slightly confusing from the examples of hybrid choice, because here the attribute importance (considered as latent variables, already have observed values.) Furthermore there are no attitudinal questions for the modelling i am trying to do.

And lastly are there output files of the examples somewhere?

Hi

1) yes, you can of course estimate a hybrid choice model sequentially too, but it is less statistically efficient to do so.

2) I don't completely understand your question. Are you looking for an example of a sequential model? Also, when you say you do not have attitudinal questions, hybrid choice models can be used for non-attitudinal indicators too of course, as in the paper you refer to, where the indicators are the stated non-attendance. The paper treats actual attribute importance as unobserved, and uses indicators of stated non-attendance as dependent variables.

We will add the outputs from model estimation of all examples to the website, but it will take us a few days, thanks for the suggestion

Stephane

Thank you for the reply, Dr Hess.

Sorry, if i was little confusing with the second question. I just have one final confusion, from what i understand the paper uses indicators of stated non-attendance as dependent variables in the choice modelling part, but uses the actual importance as the dependent variable in the initial binary logit part. Am i correct here?

If i am correct, the binary logit will help us understand what are the indicators and we will use these indicator to the hybrid model as dependent variable.

Hi

sorry, that interpretation is not quite correct.

The indicators of stated non-attendance are used as dependent variables not in the choice modelling part but in the measurement model. The link between the choice model and the measurement model is made by the latent attribute importance, which explains both the scale of parameters in the choice model and the stated attribute attendance in the measurement model

Best wishes

Stephane

Thankyou Dr Hess for the clearance,

I think my use of words were confusing, but your explanation makes it clear.