Upgrade to quantile function arising from pandas 1.3.0 internals change

.github/workflows/CI.yml

@@ -80,7 +80,7 @@ jobs:
 
  - name: Test with pytest
  shell: bash -l {0}
- run: pytest --cov=anesthetic -n auto tests
+ run: python -m pytest --cov=anesthetic -n auto tests
 
  - name: Upload coverage to Codecov
  uses: codecov/codecov-action@v1

anesthetic/utils.py

@@ -68,7 +68,7 @@ def compress_weights(w, u=None, nsamples=None):
  return (integer + extra).astype(int)
 
 
-def quantile(a, q, w=None):
+def quantile(a, q, w=None, interpolation='linear'):
  """Compute the weighted quantile for a one dimensional array."""
  if w is None:
  w = np.ones_like(a)
@@ -78,7 +78,7 @@ def quantile(a, q, w=None):
  c = np.cumsum(w[i[1:]]+w[i[:-1]])
  c /= c[-1]
  c = np.concatenate(([0.], c))
- icdf = interp1d(c, a[i])
+ icdf = interp1d(c, a[i], kind=interpolation)
  quant = icdf(q)
  if isinstance(q, float):
  quant = float(quant)

anesthetic/weighted_pandas.py

@@ -127,9 +127,11 @@ def sem(self, skipna=True):
  """Weighted standard error of the mean."""
  return np.sqrt(self.var(skipna=skipna)/self.neff())
 
- def quantile(self, q=0.5):
+ def quantile(self, q=0.5, numeric_only=True, interpolation='linear'):
  """Weighted quantile of the sampled distribution."""
- return quantile(self.values, q, self.weights)
+ if not numeric_only:
+ raise NotImplementedError("numeric_only kwarg not implemented")
+ return quantile(self.values, q, self.weights, interpolation)
 
  def compress(self, nsamples=None):
  """Reduce the number of samples by discarding low-weights.
@@ -266,16 +268,20 @@ def sem(self, axis=0, skipna=True):
  n = self.neff() if axis == 0 else self.shape[1]
  return np.sqrt(self.var(axis=axis, skipna=skipna)/n)
 
- def quantile(self, q=0.5, axis=0):
+ def quantile(self, q=0.5, axis=0, numeric_only=True,
+ interpolation='linear'):
  """Weighted quantile of the sampled distribution."""
+ if not numeric_only:
+ raise NotImplementedError("numeric_only kwarg not implemented")
  if axis == 0:
  data = np.array([c.quantile(q) for _, c in self.iteritems()])
  if np.isscalar(q):
  return Series(data, index=self.columns)
  else:
  return DataFrame(data.T, columns=self.columns, index=q)
  else:
- return super().quantile(q=q, axis=axis)
+ return super().quantile(q=q, axis=axis, numeric_only=numeric_only,
+ interpolation=interpolation)
 
  def compress(self, nsamples=None):
  """Reduce the number of samples by discarding low-weights.

tests/test_weighted_pandas.py

@@ -203,6 +203,9 @@ def test_WeightedDataFrame_quantile():
  quantile = df.quantile(qs, axis=1)
  assert isinstance(quantile, WeightedDataFrame)
 
+ with pytest.raises(NotImplementedError):
+ df.quantile(numeric_only=False)
+
 
 def test_WeightedDataFrame_hist():
  plt.figure()
@@ -405,6 +408,9 @@ def test_WeightedSeries_quantile():
 
  assert_allclose(series.quantile(qs), qs, atol=1e-2)
 
+ with pytest.raises(NotImplementedError):
+ series.quantile(numeric_only=False)
+
 
 def test_WeightedSeries_hist():
  plt.figure()