v0.17.5 (#617)

CHANGELOG.md

@@ -1,8 +1,12 @@
 ### Changelogs
 
+### 0.17.5
+ - more bugs in `plot_covariate_groups` fixed when using non-numeric strata.
+
 #### 0.17.4
  - Fix bug in `plot_covariate_groups` that wasn't allowing for strata to be used. 
  - change name of `multicenter_aids_cohort_study` to `load_multicenter_aids_cohort_study`
+ - `groups` is now called `values` in `CoxPHFitter.plot_covariate_groups`
 
 #### 0.17.3
  - Fix in `compute_residuals` when using `schoenfeld` and the minumum duration has only censored subjects. 

docs/conf.py

@@ -60,7 +60,7 @@
 #
 # The short X.Y version.
 
-version = "0.17.4"
+version = "0.17.5"
 # The full version, including dev info
 release = version
 

lifelines/fitters/coxph_fitter.py

@@ -31,7 +31,7 @@
 from lifelines.utils import (
     _get_index,
     _to_list,
-    _to_array,
+    _to_tuple,
     survival_table_from_events,
     inv_normal_cdf,
     normalize,
@@ -1580,7 +1580,7 @@ def plot_covariate_groups(self, covariate, values, **kwargs):
                 ax = plt.figure().add_subplot(1, 1, 1)
                 x_bar = self._norm_mean.to_frame().T
 
-                for name, value in zip(self.strata, _to_array(stratum)):
+                for name, value in zip(_to_list(self.strata), _to_tuple(stratum)):
                     x_bar[name] = value
 
                 X = pd.concat([x_bar] * len(values))

lifelines/statistics.py

@@ -149,14 +149,22 @@ def logrank_test(durations_A, durations_B, event_observed_A=None, event_observed
     """
     Measures and reports on whether two intensity processes are different. That is, given two
     event series, determines whether the data generating processes are statistically different.
-    The test-statistic is chi-squared under the null hypothesis.
+    The test-statistic is chi-squared under the null hypothesis. Let h_i(t) be the hazard ratio of 
+    group i at time t:
 
-     - H_0: both event series are from the same generating processes
-     - H_A: the event series are from different generating processes.
+     - H_0: h_1(t) = h_2(t)
+     - H_A: h_1(t) = k * h_2(t), k <> 1
 
 
     This implicitly uses the log-rank weights.
 
+    Important
+    ----------
+
+    The logrank test has maximum power when the assumption of proportional hazards is true. As a consquence, if the survival 
+    curves cross, the logrank test will give an inaccurate assessment of differences. 
+
+
     Parameters
     ----------
 
@@ -326,7 +334,8 @@ def multivariate_logrank_test(
     This test is a generalization of the logrank_test: it can deal with n>2 populations (and should
     be equal when n=2):
 
-     - H_0: all event series are from the same generating processes
+
+     - H_0: h_1(t) = h_2(t) = h_3(t) = ... = h_n(t)
      - H_A: there exist atleast one group that differs from the other.
 
 

lifelines/utils/__init__.py

@@ -1337,6 +1337,12 @@ def _to_list(x):
     return x
 
 
+def _to_tuple(x):
+    if not isinstance(x, tuple):
+        return (x,)
+    return x
+
+
 def format_p_value(decimals):
     threshold = 0.5 * 10 ** (-decimals)
     return lambda p: "<%s" % threshold if p < threshold else "{:4.{prec}f}".format(p, prec=decimals)

lifelines/version.py

@@ -1,4 +1,4 @@
 # -*- coding: utf-8 -*-
 from __future__ import unicode_literals
 
-__version__ = "0.17.4"
+__version__ = "0.17.5"

tests/test_plotting.py

@@ -308,11 +308,28 @@ def test_coxph_plot_covariate_groups(self, block):
     def test_coxph_plot_covariate_groups_with_strata(self, block):
         df = load_rossi()
         cp = CoxPHFitter()
-        cp.fit(df, "week", "arrest", strata=["paro"])
+        cp.fit(df, "week", "arrest", strata=["paro", "fin"])
         cp.plot_covariate_groups("age", [10, 50, 80])
         self.plt.title("test_coxph_plot_covariate_groups_with_strata")
         self.plt.show(block=block)
 
+    def test_coxph_plot_covariate_groups_with_single_strata(self, block):
+        df = load_rossi()
+        cp = CoxPHFitter()
+        cp.fit(df, "week", "arrest", strata="paro")
+        cp.plot_covariate_groups("age", [10, 50, 80])
+        self.plt.title("test_coxph_plot_covariate_groups_with_strata")
+        self.plt.show(block=block)
+
+    def test_coxph_plot_covariate_groups_with_nonnumeric_strata(self, block):
+        df = load_rossi()
+        df["strata"] = np.random.choice(["A", "B"], size=df.shape[0])
+        cp = CoxPHFitter()
+        cp.fit(df, "week", "arrest", strata="strata")
+        cp.plot_covariate_groups("age", [10, 50, 80])
+        self.plt.title("test_coxph_plot_covariate_groups_with_single_strata")
+        self.plt.show(block=block)
+
     def test_coxtv_plotting_with_subset_of_columns(self, block):
         df = load_stanford_heart_transplants()
         ctv = CoxTimeVaryingFitter()