Everyone,
Again, thanks for your insightful responses. One further question. Is it
valid to apply an ordered logit or probit model to ranked data, or must one
use exploded logistic regression? I have a scenario where 20 subjects were
asked to rank 5 symptoms in the order in which they occurred during a
migraine attack. There are ties in the data, and some symptoms were not
ranked by some subjects because they were not experienced. I would like to
test whether there is a relationship between the predicted rank order and
the observed rank order for the five symptoms. I am aware of a modified
version of the Friedman test that allows for unbalanced cells; however, this
is not implemented in Stata. Any suggestions would be greatly appreciated.
Best Regards,
Jim Shaw
BEGIN:VCARD
VERSION:2.1
N:Shaw;James;W.
FN:Shaw, James W.
ORG:The University of Arizona;College of Pharmacy
TITLE:Research Associate
TEL;WORK;VOICE:520-626-4452
TEL;HOME;VOICE:215-643-0392
TEL;CELL;VOICE:215-852-3045
TEL;WORK;FAX:520-626-3386
TEL;HOME;FAX:215-542-7150
ADR;WORK:;;P.O. Box 210207;Tucson;AZ;85710;USA
LABEL;WORK;ENCODING=QUOTED-PRINTABLE:P.O. Box 210207=0D=0ATucson, AZ 85710=0D=0AUSA
X-WAB-GENDER:2
EMAIL;PREF;INTERNET:[email protected]
REV:20030905T155602Z
END:VCARD
