ApolloChoiceModelling forum

Hi

the version without mixing should not cause problems, especially as the forthcoming version of Apollo will include analytical gradients for MDCEV and MDCNEV. For the version with mixing, you'll want to ensure you run this on a machine with sufficient RAM and a large number of cores

Stephane

If the second choice is truly only between the first pref and the opt-out, then it's indeed a sequence of two binary choices. let's say someone chooses A over B, then SQ over A. It would then be P(A|A,B)*P(SQ|B,SQ). You need to model them jointly in this way. If you only model the choice for the ...

hi

from what you describe, model B is a restricted version of model A, so the correct comparison is to use a likelihood-ratio test

Stephane

There are a number of other possible reasons here:

- class 2 captures people that do not engage with the worst choice, and the scale parameter is thus not different from 0
- the simple asymmetry assumption between best and worst might not be appropriate

On that second one, did you try separate ...

Hi

the simple logic to follow is that only differences in utility matter. So in each class, you need to make sure that your model is idenitifed, so e.g. fixing one level for categorical variables and one ASC. In the class allocation models, the characteristics are shared across classes, so there ...

Hi - you can use apollo_deltaMethod for that, just write down the expression for the difference in the alphas

Peter

how many draws are you using? And what were the runtimes like for simpler specifications of the model, e.g. just latent class? That would tell us whether there are some issues specific to your implementation, as your model per se is not beyond models we have successfully run on standard ...

Hi

yes, this clarifies things, though adding an LV does not necessarily solve endogeneity issues of course

Stephane

I would potentially interact it with the thresholds in your ordered model for car use

regarding the joint model, what is shared between the models?

Hi

I would not analyse the distribution of the conditional means alone, as that ignores the uncertainty around them. You should look at the full distribution, which you can get from the sample level results (apollo_unconditionals).

But high heterogeneity can of course make sense anyway

Stephane