Nested Logit regression. SOS!

[Carolina Bueno]

Hello everybody,

I am curretly facing two main problems in my nested logit model for explaining the type of travel pass used for riding public transport. The tree structure of a user’s decision about how to pay might look like this:

I had typed the following command:

nlogit choice ctrips || type: incom_r age_r, base(1) || mtabp:, noconstant case(id)

Firstly, I believe that INCOME, and AGE (an individual-specific variables) influence whether a user will travel without a travel card or with travel card (top level) and that the alternative-specific variable CTRIPS influence the choice of the type of travel pass at the bottom level (the 2 types of travel pass). MTABP is the variable identifying the bottom level alternatives (type=mtabp (cash:1, used pass: 2 3 )

My questions are the following:

1. How to allow an independent variable in more than one level equation. I believe that income influences in both levels. I have read that specifying the collinear option will allow execution to proceed in this case but I am not able to specify this option.

2.How is the right syntax to include categorical variables as individual specific variables. For example, with regard to age, I already included it in the command but I can not see the beta coefficients for all the age categories in the results

Thank you in advance for your valuable help!,

Paola

[Carlo Lazzaro]

Carolina (or Paola):
as far as your second question is concerned, you should include -i.- before -age_r-

i.age_r

to inform Stata you have a polytomous variable.
Otherwise, Stata considers it a continuous one.
For more on categorical variables (and interactions), just type -help fvvarlist-.

[Carolina Bueno]

Thank you, very much Carlo. 2nd Problem solved! I also had to use the xi prefix to the command for the version I am currently using.