Hello everybody,
I've been trying to run a nested logit with some data I collected in a survey. I am pretty new to this kind of model, so I might be making some mistakes.
As I checked out from past posts about nests, the problems I am facing are the following:
1) I have multiple choice (each guys interviewed chose 1 or 2 of the 5 options I proposed – 120 guys, each makes an observation with several answers). This reflects in the impossibility to define the classical "mode" variable (nlogitgen type = mode (drive:1|2, driven:3)) that defines with a number the categories chosen, i.e.:
1- rental
2- leasing
3- taxi
as people might ave chosen 1 & 2, or 1 & 3, and so on.
2) those options would be my Y, which I think it is an error because, if I got right, the Y shall be binary (but what shall it say then? I didn't get)
3) I don't know how to solve the "case" problem in the nlogit final command line, when I am specifing the last equation at the bottom level (mode:, case(?)).
–I think I could solve the first problem this way: as each replier answered several questions, I could duplicate the observations, "freeze" all the others answers and change only the Y question (when they selected 2 options) – but what to do when they picked just one?
–this could maybe lead to a solution to the third problem: as I would "duplicate" the answers, each of the guys would fall in an reference group of 2 people (he and himself replying differently), so I would solve the "case" problem, but again this would not work for who only picked one, as he would be cut out.
Did I say something wrong? If yes, please correct me, I would love to understand how to fix data properly.
Thnak you very much,
Niccolò
I've been trying to run a nested logit with some data I collected in a survey. I am pretty new to this kind of model, so I might be making some mistakes.
As I checked out from past posts about nests, the problems I am facing are the following:
1) I have multiple choice (each guys interviewed chose 1 or 2 of the 5 options I proposed – 120 guys, each makes an observation with several answers). This reflects in the impossibility to define the classical "mode" variable (nlogitgen type = mode (drive:1|2, driven:3)) that defines with a number the categories chosen, i.e.:
1- rental
2- leasing
3- taxi
as people might ave chosen 1 & 2, or 1 & 3, and so on.
2) those options would be my Y, which I think it is an error because, if I got right, the Y shall be binary (but what shall it say then? I didn't get)
3) I don't know how to solve the "case" problem in the nlogit final command line, when I am specifing the last equation at the bottom level (mode:, case(?)).
–I think I could solve the first problem this way: as each replier answered several questions, I could duplicate the observations, "freeze" all the others answers and change only the Y question (when they selected 2 options) – but what to do when they picked just one?
–this could maybe lead to a solution to the third problem: as I would "duplicate" the answers, each of the guys would fall in an reference group of 2 people (he and himself replying differently), so I would solve the "case" problem, but again this would not work for who only picked one, as he would be cut out.
Did I say something wrong? If yes, please correct me, I would love to understand how to fix data properly.
Thnak you very much,
Niccolò