[Limdep Nlogit List] best-worst data, ties and models
Daniel Rigby
dan.rigby at manchester.ac.uk
Fri Nov 6 21:37:21 EST 2009
In August 2008 Walt Paczkowski asked about data ranked best-worst, ie 4 options with the 'best' and 'worst' options identified.
Bill Greene replied:
"NLOGIT allows rank data and it allows ties. For your case, the best is ranked 1, the worst is ranked 3 and the rest are tied for second."
(He added more but I cut that here)
Later Walt asked the following question
On page N10-9, it says .that ranks data may be used with RPLOGIT. But then on page N17-23, it says that ranks data is not available for it. Which is it?
I cannot find a posted answer to this question - can someone clarify whether RPL can accommodate this data format?
I have 2 additional questions
1. Ties
Looking at NLOGIT N10-9 manual I see that models estimated on this form of data require the adding of the command:
;Ranks $
and that "The ties must only appear at the lowest level"
This seems inconsistent with Bill Greene's original reply that
"the best is ranked 1, the worst is ranked 3 and the rest are tied for second."
Can someone clarify whether Nlogit can handle the ties being the middle category, between best and worst: [1,2,2,3]
2. Items not decomposed into attributes.
I presume (but better check before I start) that the Best-Worst data used with CLOGIT and (subject to reply) possibly RPL models can be estimated using data where the choice sets comprise single items (ie sets of 4 individual items) rather than the usual case where the options comprise multiple attributes which differ in their levels.
I am interested in the utility associated with the individual item - they are not decomposed into attributes.
.
Thanks,
Dan
Dan Rigby
Environmental Economics
Manchester University
More information about the Limdep
mailing list