|
|
I have an analysis of sickness absence data in which we use survival
analysis (cox proportional hazards regression).
We are investigating which effect a particular variable has for spells of
sick leave for different durations and have a hypothesis that says that when
duration of the spells of sick-leave increase the effect of this particular
variable will decrease.
So I have a set of 11 hazard rates and want to examine whether there is a
linear relationship between these estimates of the rates with the number of
weeks that is used as 'cut-off'-point for. My first suggestion was just to
use pearson correlation coefficient to test for a linear relationship.
However, if I only do that I do not take into account the uncertainty of the
hazard rates (i.e. I do not use the information on the confidence intervals
or the standard error). This is particularly important because the number of
events used in the survival analysis decrease when we increase the cut-off
point for how many weeks the spells of sick-leave has to be in order for
them to be included in the analysis. I don't know if you can use the
standard error for the hazard rates in an analysis or you should instead use
the number of events that the hazard rate is based upon?
The data looks something like this:
Number of weeks Hazard rate Number of respondents
2 1.25 1272
3 1.22 1038
4 1.21 878
. .
20 0.95 192
For each hazard rate I also have the standard error associated with the
estimate.
Does anybody have any suggestions on what to use and whether it can be
carried out in SPSS?
Thank you very much in advance,
Claus D. Hansen
|
Locked
