[R] Error in check(itp) : ‘object’ does not represent a K sample problem with censored data

[Bob Green]

Hello,

I have two questions regarding a survival analysis I have been 
working on. Below is the code to date.
The variables:

1) recidivism$intDaysUntilFVPO  are the number of days before an 
violent offence was committed - if no offence was committed than the 
days between court hearing and end of data collection was recorded.
2) recidivism$intDaysUntilFNVPO are the number of days before a 
nonviolent offence was committed - if no offence was committed than 
the days between court hearing and end of data collection was recorded.
3) recidivism$Event_v   coded TRUE or FALSE as to whether a violent 
offence was committed
4) recidivism$Event_nv   coded TRUE or FALSE as to whether a 
non-violent offence was committed
5) recidivism$group - there were 3 groups recorded as 1, 2 or 3

I encountered two problems:

a)  when I tried to run the following: > 
surv_test(Surv(recidivism$intDaysUntilFVPO, recidivism$Event_v) ~ 
recidivism$group, data = recidivism, distribution = "exact")

I received the following error message. I am unsure whether the 
problem lies with the group variable, how I have set up the time 
variables or there is something missing.

Error in check(itp) :
   'object' does not represent a K sample problem with censored data

(b) I ran some simple plots of the curves. While I have a fair idea 
of  which group is which, I wanted to add a legend identifying the 
groups ? What code is required for this?

Any assistance is much appreciated,

Bob


 > library(survival)
 >
 > survdiff(Surv(recidivism$intDaysUntilFVPO, recidivism$Event_v) ~ 
group, data = recidivism)
Call:
survdiff(formula = Surv(recidivism$intDaysUntilFVPO, recidivism$Event_v) ~
     group, data = recidivism)

            N Observed Expected (O-E)^2/E (O-E)^2/V
group=1 1136      135   154.18      2.39      7.75
group=2  446       87    61.29     10.78     14.90
group=3   65        1     7.52      5.66      5.86

  Chisq= 18.9  on 2 degrees of freedom, p= 8.06e-05
 > survdiff(Surv(recidivism$intDaysUntilFNVPO, recidivism$Event_nv) ~ 
group, data = recidivism)
Call:
survdiff(formula = Surv(recidivism$intDaysUntilFNVPO, recidivism$Event_nv) ~
     group, data = recidivism)

            N Observed Expected (O-E)^2/E (O-E)^2/V
group=1 1136      323    366.3      5.11     17.80
group=2  446      184    127.7     24.84     33.12
group=3   65        7     20.1      8.51      8.86

  Chisq= 38.5  on 2 degrees of freedom, p= 4.3e-09
 > layout(matrix(1:2, ncol = 2))
 > plot(survfit(Surv(recidivism$intDaysUntilFNVPO, 
recidivism$Event_v) ~ recidivism$group, data = recidivism), main = 
"time to violent reoffense by group", lty = c(1,2,3), ylab = 
"probability", xlab = "survival time in days")
 > plot(survfit(Surv(recidivism$intDaysUntilFVPO, 
recidivism$Event_nv) ~ recidivism$group, data = recidivism), main = 
"time to nonviolent reoffense by group", lty = c(1,2,3), ylab = 
"probability", xlab = "survival time in days")
 >
 > library(coin)
 > surv_test(Surv(recidivism$intDaysUntilFVPO, recidivism$Event_v) ~ 
recidivism$group, data = recidivism, distribution = "exact")
Error in check(itp) :
   'object' does not represent a K sample problem with censored data