Lifelines

Section 13.7 Lifelines

A Python package called lifelines provides tools for survival analysis, including functions that compute Kaplan-Meier estimates.

🔗

We can use it to confirm that the result in the previous section is correct. First we’ll compute the survival function using estimate_survival.

🔗

Listing 13.7.1. Python Code

🔗

try:
    import lifelines
except ImportError:
    %pip install lifelines

Listing 13.7.2. Python Code

🔗

surv = estimate_survival(resp60)

Next we’ll compute it using lifelines. First we’ll get the data into the format lifelines requires.

🔗

Listing 13.7.3. Python Code

🔗

complete = complete.dropna()
durations = np.concatenate([complete, ongoing])
event_observed = np.concatenate([np.ones(len(complete)), np.zeros(len(ongoing))])

Now we can make a KaplanMeierFitter object and fit the data.

🔗

Listing 13.7.4. Python Code

🔗

from lifelines import KaplanMeierFitter

kmf = KaplanMeierFitter()
kmf.fit(durations=durations, event_observed=event_observed)

<lifelines.KaplanMeierFitter:"KM_estimate", fitted with 15389 total observations, 5468 right-censored observations>

After fitting the data, we can call the plot function to display the results, which include the estimated survival function and a confidence interval—although the confidence interval is not correct in this case because it doesn’t correct for stratified sampling.

🔗

Listing 13.7.5. Python Code

🔗

kmf.plot()

decorate(xlabel="Age", ylabel="Prob never married", ylim=ylim)

Unlike the survival function we computed, the one from lifelines starts from 0. But the rest of the function is the same, within floating-point error.

🔗

Listing 13.7.7. Python Code

🔗

ps = kmf.survival_function_["KM_estimate"].drop(0)
np.allclose(ps, surv)

True

In the next section, we’ll use weighted resampling to compute confidence intervals that take stratified sampling into account.

🔗

Prev Top Next