nested logit about cell phone

classic Classic list List threaded Threaded
1 message Options
Reply | Threaded
Open this post in threaded view
Report Content as Inappropriate

nested logit about cell phone

This post has NOT been accepted by the mailing list yet.
This post was updated on .

I have conducted a pet project about smartwatches and collected some data. I ran a clogit and am trying to do a nested logit model. I haven't done a nlogit before. Thus, I am not sure how to calculate the probabilities of selecting the alternatives at the bottom level and the probabilities of selecting the branches on the top level. If someone could help I would really appreciate it!

First, here is my 2 levels tree structure

Here is my nlogit codes:
nlogit choice <all alternative specific variables> || type: platforminterest, base (none)|| alt:, noconstant case(groupid) nolog

Please note that platforminterest is a case-specific variable with values 1, 2, 3 where 1 means currently using android platform for their cell phone, 2 means using ios platform, 3 means not interested in buying at all.

Below is the output, I do notice that many of the coefficients are not significant. That's because I have less than 20 responses...anyway, let's assume all coefficients are significant for the sake of illustration...

- What are the coefficients in 'type equations' used for?
- How can I calculate the probabilities of 4 alternatives and the probabilities of selecting each branch?

I also asked Stata to predict:
- p1: probabilities of selecting each branch at the top level
- p2: probabilities of selecting each alternative at the bottom level
- condp: conditional probabilities of selecting an alternative given a branch is chosen
- xb1: linear prediction at the top level - same as the type equation coefficients in the output
- xb2: linear prediction at the bottom level
- iv: inclusive value

I know how to calculate xb2. And I know that p2= p1*condp. But I am not sure how to calculate p1 and condp. And what is the role of iv? Any advice would be helpful. Thanks!