Bookmark and Share

Notice: On April 23, 2014, Statalist moved from an email list to a forum, based at statalist.org.


[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]

st: nested logit and discrete choice data


From   Nils Wlömert <[email protected]>
To   [email protected]
Subject   st: nested logit and discrete choice data
Date   Mon, 29 Nov 2010 11:44:31 +0100

Dear listers,

I am trying to define a nested logit model for discrete choice data derived from a conjoint choice experiment, where participants chose between 3 different product alternatives plus a base alternative "none of these products" in 8 consecutive choice decisions. I would like to nest the base "no-choice" option separate from the "real" alternatives very much like Haaijer, Kamakura and Wedel (2001) in their paper: http://faculty.fuqua.duke.edu/~kamakura/My%20Reprints/The%20no-choice%20alternative%20in%20conjoint%20choice%20experiments.pdf

That is, the first level of the decision tree is the choice between one of the real alternatives (1,2,3) vs. the no choice option (4) (captured in the dissimilarity IV parameter) and the second (bottom) level is the standard MNL choice probability depending on the product attributes within the "real choice" nest. Products are described by 6 attributes with 4-5 levels each. Hence, there are 6 columns in the data set indicating the levels of each attribute (1-5) and a 1,0 dummy for the no-choice option (every 4th row). The dependent variable "decision" is also a 1,0 dummy. Further, there are columns for alternativeID (32 alternatives per respondent), choice-setID (8 sets per respondent), taskID (1,2,3 for the real alternatives and 4 for the no-choice option) and caseID. I generated a new variable from taskID using nlogitgen including 1 "choice" (1,2,3) and 2 "nochoice" (4).

I searched the documentation and web resources (e.g., the paper on nested logit by Florian Heiss http://www.stata-journal.com/sjpdf.html?articlenum=st0017) . However, I find that this is different from the given examples in that the independent variables are not the levels of one specific variable (e.g.: restaurant type, travel mode), but rather the levels of several different variables (being the 4-5 levels of the 6 attributes in 6 different rows in the data set). I would be interested to learn, how to define a model that accomodates this kind of data structure? What would be my variables at the two levels of the hierarchy?

Any help is much appreciated!

Thanks & best,

Nils.

*
*   For searches and help try:
*   http://www.stata.com/help.cgi?search
*   http://www.stata.com/support/statalist/faq
*   http://www.ats.ucla.edu/stat/stata/


© Copyright 1996–2018 StataCorp LLC   |   Terms of use   |   Privacy   |   Contact us   |   Site index