Competing-risks regression

[Gaston Fernandez]

Hi all!

I am having issues setting up and estimating a competing-risk regression using the command stcrreg. Any help on this is much appreciated.

I have an unbalanced panel of individuals. Individuals transit over time through different marital statuses (i.e., singlehood, cohabitation, and marriage). The data looks like this:

clear
input str7 id float(year marital_status) int age
"1" 2002 0 25
"1" 2005 2 27
"1" 2010 2 32
"2" 2002 0 30
"2" 2003 0 31
"2" 2004 1 32
"2" 2005 1 33
"2" 2006 2 34
"2" 2010 2 43
"2" 2011 0 44
"3" 2010 0 26
"4" 2002 0 40
"4" 2004 2 42
"4" 2006 1 44
"4" 2008 0 46
end

where the variable marital_status is coded as:

label define status 0 "singlehood" 1 "cohabitation" 2 "marriage"

label values marital_status status

tab marital_status, m

marital_stat |
          us |      Freq.     Percent        Cum.
-------------+-----------------------------------
  singlehood |          7       46.67       46.67
cohabitation |          3       20.00       66.67
    marriage |          5       33.33      100.00
-------------+-----------------------------------
       Total |         15      100.00

I am interested in studying the impact of age on the likelihood that a given marital status occurs. For that, I am trying to estimate sub-hazard ratios between different competing marital statuses but apparently, I am not setting up the data correctly since I get the following error message:

option compete(): competing risks events must be stset as censored

For instance, for the relative likelihood to cohabit (vs being married), I am using the following:

clonevar year1 = year

snapspan id year1 marital_status, gen(year0) clear

stset year1, id(id) origin(time year0) failure(marital_status = 2)

Survival-time data settings

           ID variable: id
         Failure event: marital_status==2
Observed time interval: (year1[_n-1], year1]
     Exit on or before: failure
     Time for analysis: (time-origin)
                Origin: time year0

--------------------------------------------------------------------------
         15  total observations
          1  ignored because never entered
          3  observations end on or before enter()
          5  observations begin on or after (first) failure
--------------------------------------------------------------------------
          6  observations remaining, representing
          3  subjects
          3  failures in single-failure-per-subject data
          9  total analysis time at risk and under observation
                                                At risk from t =         0
                                     Earliest observed entry t =         0
                                          Last observed exit t =         4

 

stcrreg age, compete(marital_status = 1) robust
option compete(): competing risks events must be stset as censored

Any ideas?
Thanks!

[Stata version 17.0]

[Paul Dickman]

Try with double equal signs in both stset and stcrreg.

stset year1, id(id) origin(time year0) failure(marital_status == 2)

   
 stcrreg age, compete(marital_status == 1) robust

[Gaston Fernandez]

Thanks, Paul. Double-equal signs do not make the trick. I think the error message is related to the way I define failures and non-failures (i.e., censored observations) in the stset command.