cph.plot_partial_effects_on_outcome()

[alexandre-sousa-git]

Hello ,

I'm new with this library , and i have a question about this function, is it possible extract the values plotted by the function?

[CamDavidsonPilon]

Hi @alexandre-sousa-git - no, there isn't, but this has been brought up before. Are you interested in the the baseline survival line too?

[CamDavidsonPilon]

That's right

[alexandre-sousa-git]

What I really needed was to know the coordinates of each curve of each covariate not the baseline. For example the difference of the survival for each gender .

[CamDavidsonPilon]

I mean, two ways to do that:

1. Look at the regression coefficient on gender with .print_summary() - that provides a summary of the survival differences
2. Manually determine the survival curves:

cph = CoxPHFitter().fit(rossi, "week", "arrest")

baseline = cph._central_values.copy()
cph.predict_survival_function(baseline) # baseline

X1 = baseline.copy()
X1['baseline', 'arrest'] = 1
cph.predict_survival_function(X1) # alternative

[alexandre-sousa-git]

Thanks a lot! This will help! congrats for your work

[dlovell]

@CamDavidsonPilon : I believe this snippet will extract the data from the AxesSubplot returned by plot_partial_effects_on_outcome

pd.DataFrame({
    line.get_label(): pd.Series(
        line.get_ydata(),
        line.get_xdata(),
    )
    for line in ax.lines
})