Each patient can be in or out of a particular cohort. With 5 cohorts
that gives 2^5 = 32 possible histories. Some of them may not be
observed. Even if they all occur some kind of bar chart should still be
possible. This could show
1. The tallies over histories: so the categories could be 1, 12, 123,
1234, 12345, 13, etc. In principle that keeps all the information.
2. The tallies in each cohort, subdivided by when the patient first
joined the study. That throws away some information.
Nick
[email protected]
Jeph Herrin
This is not a Stata question per se, but I hope the
solution is something I can do in Stata.
I have a dataset consisting of 5 overlapping cohorts
of patients. Basically, every 6 mos data is collected
on every (diabetes) patient seen at a network of clinics.
I have the first 2.5 years (there will be another 2.5).
There is a lot of overlap between the 5 cohorts. There are
about 10k obs per cohort, and each "snapshot" includes about
80% of patients that are seen in another cohort. I would
like to illustrate this somehow, because the amount of
overlap is of interest. The figure should also capture the
time trend (ie, a Venn diagram wouldn't really do it).
I imagine some sort of figure with 5 bars, each with height
representing the # obs and divided somehow to represent the
number of patients in that cohort that were seen in other
cohorts. Ideally - and this is the hard part - the figure
would also indicate somehow the number of times patients in
that cohort were seen in other cohorts.
I'm just fishing here for ideas - I've never seen anything
like this, but maybe some of our talented listers have?
*
* 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/
