Talk:Mixed logit

From DDL Wiki

Norman,

This page is mixing up the concept of the form of the utility function for an individual consumer (continuous vs. discrete) with the concept of the distribution of the utility function parameters across the consumers. Mixed logit deals with a distribution of beta values (continuous or discrete) over the consumers, whereas the graphs you have drawn show utility as a function of the attribute level. These are two different concepts. I hope that the discussion we had in the lab today has cleared this up. If not, please ask.

Jeremy,

You are right. I explained the things in the way you described but the figures could make people confused. I just seperated figure into three and will add more explaination. Norman

Norman, I still think this is confusing listing them as three cases a), b), c). Really, case c) is the random coefficients version of a). Move the discussion of a) and b) to a different page about forms of utility functions for random utility models, and for this page, you can make a single example (suppose v = beta*price) and show the difference for that example between standard and mixed logit. If you want, you could also show what a mixed logit would look like in the discrete case, but this is a more complex example, and the discrete/continuous is a separate issue alltogether than mixed logit.

..