I have data from a study where subjects were shown videos and asked to identify the contents as one of two options, as well as to rate their confidence of correct classification on a likert scale of 1 to 7. Each subject was asked to perform this classification task multiple times, and given different video material each time.

I am using mixed logit because I think this data is categorical. I am encoding it as each choice situation having 14 alternatives.

That is, the subject identifies the content as type 1 or type 2 then provides a rating of their confidence from 1 to 7, so I view that as 2 * 7 = 14 alternatives of which only 1 may be given as the answer.

Does this seem a reasonable approach? Or is this an inappropriate way to use mixed logit estimation?

I'm new to discrete choice methods, and my general statistical knowledge is rusty, so I welcome reading recommendations. Thanks very much for your time!


1 Answer 1


The first thing you would like to do is to decide whether you going to treat your data as a 14 levels from 'totally sure negative opinion' to 'totally sure positive opinion', or as nested choice (which seems more natural given your questioning design) - first between negative and positive option, and then within each of these two options between another seven options.

Essentially such models with sequential choice are called nested logit models, but there is an analogue of such nested models for mixed logit simulations (see, e.g., section 6.3 of this: elsa.berkeley.edu/choice2/ch6.pdf).

  • $\begingroup$ I think you are right that I need something nested. I just want to check that I understand the approach you linked to: Could the dummy variable be an additional independent variable which is 1 if the participant has answered correctly and 0 if the participant was incorrect? With a dependent variable that still has 14 levels ranging from correct answer plus highest confidence rating to incorrect answer plus highest confidence rating? Or do I misunderstand? Thanks very much! (And thanks for the recommendation of Train's book! It's great!) $\endgroup$
    – user43721
    Apr 17, 2014 at 23:49

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Not the answer you're looking for? Browse other questions tagged or ask your own question.