Hello everyone,
A question on the suitability of a specific data design for an ordinal
logistic regression...
Say we have 10 judges, each giving an opinion (rated
1=poor...5=excellent) for each of 14 items. Each judge rates the items
independently of other judges. We thus end up with 140 responses.
We wish to relate 'item description' to response. Our data may look
like (hypothetical example):
Item judge colour fabric response
1 1 blue cotton 5
1 2 blue cotton 4
1 3 blue cotton 4
. .
. .
. .
1 10 blue cotton 2
. .
. .
14 1 red leather 3
14 2 red leather 2
14 3 red leather 4
. .
. .
14 10 red leather 3
We wish to produce a model relating item description to response.
Obviously item description will be approriately coded using dummy
variables.
Since the response is ordinal, ordinal logistic regression appears to be
a suitable modelling technique.
I would like to ask, is it appropriate in the scenario described above
to enter data as a Nxp matrix (where N=140 in this hypothetical
example).
Many thanks for your views on this,
Kind Regards,
Kim
|