choice-metrics.com

Dear all,

I was wondering if it could be necessary to use stochastic variables instead of deterministic variables (for the attributes or covariates) in my Mixed Logit model/Latent Class model.
I did not find pretty much literature on this topic besides "Diaz et al, 2015, Accounting for stochastic variables in discrete choice models". Maybe you could give any advice on when to use stochastic instead of deterministic specified variables or when / how to test for it.

Thank you very much in advance.
Best,
Juli

In Diaz et al they seem to mainly consider revealed preference (RP) data as they are talking about perception errors, i.e. differences between the actual attribute level and the perceived/reported attribute level. In stated choice experiments, there is not really any perception error because all attribute levels are shown to the respondents, so I do not see the need to use stochastic variables in the utility function when using SP data.

Michiel