I would like to know the difference in my survival analysis approach.
I have 4 types of events that I want to examine, and not everyone experiences an event.
These events are dummies in my dataset.
I use the log-logistic approach, and I consider one event only, so the other events are censored, and the rest is therefore also treated as censored.
(1) My question is then, what am I really examining when I do this for all the four event types separately? Because I don't understand the difference then when using competing risks. It looks like I am already using a form of competing risks when examining the event types separately, although without using the specific command, is this correct? I mean when running the regression with only event type 1, event type 2 3 4 and the rest is treated as censored, as in the competing risk regression, right?
Is it just with competing risks that I consider event type 1 versus event type 2, and thereby consider event type 3 and 4 and the others as censored. So the results will only involve event type 1 and 2?
Is my approach, regarding the first question, meaningful?
Or should I run the log-logistic regression only for all the event types together, and use the competing risks approach for the comparison of the event types?
(2) My second question involves the model distribution when using a competing risks approach. I read some papers that mention 'a log-logistic AFT competing risks model', but how should I determine the distribution for the competing risks regression? I don't see this option in the stcrreg command window.
Thanks in advance.
Kind regards,
Michael
Commands:
Separate log-logistic analysis:
Code:
stset E_Date, failure(Event==1) id(ID) enter(Date1) origin(Date1) streg $xlist, dist(loglogistic) stset E_Date, failure(Event==2) id(ID) enter(Date1) origin(Date1) streg $xlist, dist(loglogistic) stset E_Date, failure(Event==3) id(ID) enter(Date1) origin(Date1) streg $xlist, dist(loglogistic) etc.
Code:
stset E_Date, failure(Event==1) id(ID) enter(Date1) origin(Date1) stcrreg $xlist, nohr compete(Event==2) offset(Event==1)
0 Response to Competing risks in survival analyses
Post a Comment