[R] error in survival analysis

Terry Therneau therneau at mayo.edu
Tue Jul 26 14:43:54 CEST 2011

 There is something very odd about your data to give se values that are
so very large.  Usually, this means that the data is deterministic: some
combination of predictors is able to separate a subset with no events
from all the others.  This leads to a problem where the log-likelihood
is maximum at infinity.  
  I expect that both the warning messages you see are completely

Terry Therneau

---- begin included message ---
This is a simple R program that I have been trying to run. I keep
running into the "singular matrix" error. I end up with no sensible
results. Can anyone suggest any changes or a way around this?

I am a total rookie when working with R.


> library(survival)
Loading required package: splines
> args(coxph)
function (formula, data, weights, subset, na.action, init, control,
    method = c("efron", "breslow", "exact"), singular.ok = TRUE,
    robust = FALSE, model = FALSE, x = FALSE, y = TRUE, tt, ...)
test1<-read.table("S:/FISHDO/03_Phase_I_Field_Work/Data_6_28_2011/Working Folder/R_files/4SondesJuly24.csv", header=T, sep=",")
> sondes<-coxph(Surv(Start, Stop, Depart)~DOLoomis + DOI55 + DODamen,
Warning messages:
1: In fitter(X, Y, strats, offset, init, control, weights = weights,  :
  Loglik converged before variable  1,2 ; beta may be infinite.
2: In coxph(Surv(Start, Stop, Depart) ~ DOLoomis + DOI55 + DODamen,  :
  X matrix deemed to be singular; variable 3
> summary(sondes)
coxph(formula = Surv(Start, Stop, Depart) ~ DOLoomis + DOI55 +
    DODamen, data = test1)

  n= 1737, number of events= 58
   (1 observation deleted due to missingness)

               coef  exp(coef)   se(coef)  z Pr(>|z|)
DOLoomis -2.152e+00  1.163e-01  1.161e+05  0        1
DOI55     4.560e-01  1.578e+00  3.755e+04  0        1
DODamen          NA         NA  0.000e+00 NA       NA

         exp(coef) exp(-coef) lower .95 upper .95
DOLoomis    0.1163     8.5995         0       Inf
DOI55       1.5777     0.6338         0       Inf
DODamen         NA         NA        NA        NA

Concordance= 0.5  (se = 0 )
Rsquare= 0   (max possible= 0.01 )
Likelihood ratio test= 0  on 2 df,   p=1
Wald test            = 0  on 2 df,   p=1
Score (logrank) test = 0  on 2 df,   p=1

More information about the R-help mailing list