detect coxph ill fits

[dchiu911]

Is there a method to detect ill-fitted coxph models such as when these warnings appear

Warning message:
In coxph.fit(X, Y, istrat, offset, init, control, weights = weights,  :
  Loglik converged before variable  1 ; coefficient may be infinite. 

or is the only way to do this via a tryCatch() for warning() since the model object doesn't store this phenomenon? I noticed that is_converged() only has methods for mixed models.

[bwiernik]

Can you share a reproducible example that returns this warning ?

[dchiu911]

library(survival)
lung$tmp <- factor(c(rep(0, 227), 1), levels = c(0, 1))
coxph(Surv(time, status) ~ tmp, data = lung)
#> Warning in coxph.fit(X, Y, istrat, offset, init, control, weights = weights, :
#> Loglik converged before variable 1 ; coefficient may be infinite.
#> Call:
#> coxph(formula = Surv(time, status) ~ tmp, data = lung)
#> 
#>            coef  exp(coef)   se(coef)      z     p
#> tmp1 -1.401e+01  8.248e-07  1.983e+03 -0.007 0.994
#> 
#> Likelihood ratio test=0.61  on 1 df, p=0.4329
#> n= 228, number of events= 165

^{Created on 2025-07-11 with reprex v2.1.1}