Potential issue in user guide for CoxnetSurvivalAnalysis

[mfergie]

From my understanding of the user guide on CoxnetSurvivalAnalysis, the GridSearchCV should be evaluated over the set of alphas identified by fitting the CoxNet model to the full data. Instead, it is ran over the set of alphas defined using linspace further up in the code.

Unless I'm mistaken, the code in box 11 should be:

cv = KFold(n_splits=5, shuffle=True, random_state=0)
gcv = GridSearchCV(
    make_pipeline(StandardScaler(), CoxnetSurvivalAnalysis(l1_ratio=0.9)),
    param_grid={
        "coxnetsurvivalanalysis__alphas": [
            [v] for v in coxnet_pipe.named_steps["coxnetsurvivalanalysis"].alphas_
        ]
    },
    cv=cv,
    error_score=0.5,
    n_jobs=4).fit(Xt, y)

cv_results = pd.DataFrame(gcv.cv_results_)

This gives different results:

P.s. thanks for all your work on this library.

[sebp]

Good catch! It explains why the graph looked so weird.