st: From: Page Liva <page.liva@gmail.com>

[owner-statalist@hsphsun2.harvard.edu]

Dear all

this is not (yet) a Stata question but I would very much appreciate if
you could share your ideas with me. I'm looking for a sound and
practical method to summarize ranking-like data and have been reading
and searching for possible solutions for quite a long time now but
without much success. Here is my problem:
In a survey 500 patients were asked to state the top three symptoms
they suffer most of. It was NOT a list of symptoms to be selected and
ranked, but patients were asked to name the symptoms themselves.
The responses are for example:
patient A: 1) depression 2) nausea 3) cough
patient B. 1) pain 2) depression 3) rush

The 900 responses received (not every patient returned three
symptoms), were classified into 12 categories (by two raters with high
reliability but this is irrelevant).As a result we have the frequency
per category together with the severity ranks. For example:

Cat 1: 200 (severity 1), 42 (severity 2), 10 (severity 3); total count: 252
Cat 2: 110 (severity 1), 76 (severity 2), 23 (severity 3); total count: 209
Cat 3: 70 (severity 1), 80 (severity 2), 2 (severity 3); total count: 152
 ……
 In other words, we have partial rankings of categories of a variable item list.
I am now thinking about some form of summary, aggregation measure or
index for the data.
This measure should possibly reflect the (relative) frequency by which
the categories occurred, and the (average) ranking of this category.
I have considered measures of consensus which provide interesting
information but is not exactly what I want.
Do you have any creative suggestions or recommendations on how to proceed?
Many thanks in advance for your help!

Best wishes

Dean

*
*   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/