Dear All,
I apologise if this is too simplistic but my head is going around in
circles at the moment. I have data on 500+ patients (some survivors and sone
non0survivors) that is the proportion of their hospital admission (which
were for different lengths of time) that they had a blood sugar within a
predefined range of values e.g. <4, 4-6, 6-8, 8-9, >10 (all my intervals are
not equal in size). Therefore all patients have a reading in each interval
which may range between 0 and 1. I want to test my theory that non-survivors
in general spent a higher proportion of their admission with higher blood
sugars. The graphs certainly look like this but how do I test for this ? My
concerns are that I am testing a proportion and I don;t think that my groups
are independent of one another (since if a patient has a value of 1 in one
interval then all the other intervals must be zero as they must sum up to
1).
Thanks
Simon Finney
|