12-02-2014 07:28 PM
I am looking to compare my output for a treatment effect in a regression model to an estimate of crude event rate obtained without using a regression model. The regression model is not causing any difficulties, but my other model is. I want to compute my total number of events and the follow-up time for each of my treatment groups so that I can provide a crude event rate for each of my treatment groups, but thus far I have not been able to achieve this. I assumed I could use a proc freq procedure:
12-02-2014 07:44 PM
How would you calculate this by hand?
Its easiest if you post sample data (not necessarily actual data) with the desired output format/calculation.
For time based analysis I'm assuming you're using proc phreg? or proc lifereg?
12-02-2014 08:18 PM
To calculate it by hand, I would only need a summary of the overall incidence between my treatment groups. Then I could provide a estimate of the crude event rate for each group. My data would look similar to this, except there will be many more observations:
Id trtgrp nevents followuptime
1 1 3 5
2 1 4 6
3 1 3 4
4 0 5 5
5 0 7 5
6 0 6 5
7 0 6 4
12-02-2014 09:08 PM
What would the output be for this particular data example?
I'm just confused with an ID and the the nevents being 3, indicates multiple events per ID?