Jupyter Notebook Viewer

	treatment	time	failure
37	1	12	1
38	1	15	1
39	1	17	1
40	1	22	1
41	1	23	1

In [36]:

data = data.dropna()
print data.head()

   treatment  time  failure
0          2     6        1
1          2     6        1
2          2     6        1
3          2     6        0
4          2     7        1

In [37]:

treatment = data['treatment'] == 2

In [38]:

T = data['time']
E = data['failure']
t = np.linspace(0,40,150)

In [39]:

kmf = KaplanMeierFitter()
kmf.fit(T[treatment], timeline=t, event_observed=E[treatment], label='With treatment')
ax = kmf.plot()

kmf.fit(T[~treatment], timeline=t, event_observed=E[~treatment], label="Without treatment")
kmf.plot(ax=ax)
ylim(0,1.05)

Out[39]:

(0, 1.05)

Hazards, too¶

In [40]:

naf = NelsonAalenFitter()

In [41]:

naf.fit(T[treatment],timeline=t, event_observed=E[treatment], label="With treatment")
ax = naf.cumulative_hazard_.plot()
naf.fit(T[~treatment], timeline=t, event_observed=E[~treatment], label="Without treatment")
ax = naf.cumulative_hazard_.plot(ax=ax)

Try regression models¶

In [42]:

from lifelines import AalenAdditiveFitter

In [43]:

aaf = AalenAdditiveFitter()
aaf.fit(data, duration_col='time', event_col='failure', timeline=t)

 [-----------------100%-----------------] 30 of 30 complete in 0.0 sec

Out[43]:

<lifelines.AalenAdditiveFitter: fitted with 42 observations, 12 censored>

In [49]:

aaf.plot()

Out[49]:

<matplotlib.axes._subplots.AxesSubplot at 0x111606790>

In [45]:

from lifelines import CoxPHFitter

In [46]:

cp = CoxPHFitter()
cp.fit(data, duration_col='time', event_col='failure')

Out[46]:

<lifelines.CoxPHFitter: fitted with 42 observations, 12 censored>

In [47]:

cp.print_summary()

n=42, number of events=30

             coef  exp(coef)  se(coef)       z      p  lower 0.95  upper 0.95     
treatment -1.5092     0.2211    0.4096 -3.6849 0.0002     -2.3121     -0.7063  ***
---
Signif. codes:  0 '***' 0.001 '**' 0.01 '*' 0.05 '.' 0.1 ' ' 1 

Concordance = 0.690