I am planning a survey that will ask respondents to evaluate several
different clinical scenarios and, in each scenario, choose a treatment from
a set of 4 alternatives. I anticipate that I will need to use nlogit to
analyze the results, as some treatment options are more similar than others
and IIA will likely be violated.
The nlogit documentation seems to consider the case when each respondent
(e.g., family) makes only one choice (e.g., a restaurant). However, in my
situation each respondent will make one choice for each of several different
scenarios, and these scenarios will be composed of varying attributes that
will be some of the main variables of interest for analysis. The
respondents themselves will also have characteristics that will be targets
of the analysis (and, obviously, will not vary across scenarios for that
respondent).
Any suggestions on how to set up the data to achieve this? I could create
one row of data for each potential choice for for each respondent and repeat
this for each scenario, but I would then need to account for correlations
between answers given by the same respondent (cluster option?). Any other
ideas?
Thanks in advance.
Hari Nathan, MD
House Staff, Department of Surgery
Fellow, Clinical Research Scholars Program
The Johns Hopkins University School of Medicine
*
* 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/