Cannot predict survival function

[hminle]

Hi there,

I got an error when using this code:
(coxph is CoxPHSurvivalAnalysis)

surv = coxph.predict_survival_function(X_test, return_array=True)

for i, s in enumerate(surv):
    plt.step(coxph.unique_times_, s, where="post")
plt.ylabel("Survival probability")
plt.xlabel("Time in months")
# plt.grid(True)
plt.show()

I am not sure what happens.
Could you please help?

Thank you for the amazing package.

[sebp]

The -1 seems to be the problem. Does coxph.unique_times_ contain the -1?

[hminle]

Hi @sebp , yes it does

[sebp]

Can you please post which version you are using (see the issue template).

[hminle]

Hi @sebp
Here is my system

I do not have cvxopt and cvxpy

System:
    python: 3.11.4 (main, Jul  5 2023, 13:45:01) [GCC 11.2.0]
executable: /home/hminle/miniconda3/envs/py311/bin/python
   machine: Linux-5.4.0-84-generic-x86_64-with-glibc2.31

Python dependencies:
      sklearn: 1.2.2
          pip: 23.2.1
   setuptools: 68.0.0
        numpy: 1.24.3
        scipy: 1.11.1
       Cython: None
       pandas: 2.0.3
   matplotlib: 3.7.1
       joblib: 1.2.0
threadpoolctl: 2.2.0

Built with OpenMP: True

threadpoolctl info:
       filepath: /home/hminle/miniconda3/envs/py311/lib/libmkl_rt.so.2
         prefix: libmkl_rt
       user_api: blas
   internal_api: mkl
        version: 2023.1-Product
    num_threads: 32
threading_layer: intel

       filepath: /home/hminle/miniconda3/envs/py311/lib/libiomp5.so
         prefix: libiomp
       user_api: openmp
   internal_api: openmp
        version: None
    num_threads: 32

       filepath: /home/hminle/miniconda3/envs/py311/lib/libgomp.so.1.0.0
         prefix: libgomp
       user_api: openmp
   internal_api: openmp
        version: None
    num_threads: 64
sksurv: 0.21.0
numexpr: 2.8.4
osqp: 0.6.3

[sebp]

I can reproduce this issue if y passed to fit contains an event with time -1. This should actually already thrown an exception during fit, because time points must be non-negative.

See #406

[hminle]

Thank you so much @sebp . I got my bug