def test_negative_values_in_weights_error(self):
"""Test that an exception is raised if there are negative values in sample_weight."""
x = CappingTransformer(capping_values={"a": [2, 10]})
with pytest.raises(ValueError, match="negative weights in sample weights"):
x.weighted_quantile([2, 3, 4, 5], [0, 1], [2, -0.01])
def test_zero_total_weight_error(self):
"""Test that an exception is raised if the total sample weights are 0."""
x = CappingTransformer(capping_values={"a": [2, 10]})
with pytest.raises(
ValueError, match="total sample weights are not greater than 0"
):
x.weighted_quantile([2, 3, 4, 5], [0, 1], [0, 0])
def test_inf_values_in_weights_error(self):
"""Test that an exception is raised if there are inf values in sample_weight."""
x = CappingTransformer(capping_values={"a": [2, 10]})
with pytest.raises(ValueError, match="infinite values in sample weights"):
x.weighted_quantile([2, 3, 4, 5], [0, 1], [2, np.inf])
with pytest.raises(ValueError, match="infinite values in sample weights"):
x.weighted_quantile([2, 3, 4, 5], [0, 1], [1, -np.inf])
def test_expected_output(
self, values, sample_weight, quantiles, expected_quantiles
):
"""Test that weighted_quantile gives the expected outputs."""
x = CappingTransformer(capping_values={"a": [2, 10]})
values = pd.Series(values)
actual = x.weighted_quantile(values, quantiles, sample_weight)
# round to 1dp to avoid mismatches due to numerical precision
actual_rounded_1_dp = list(np.round(actual, 1))
assert (
actual_rounded_1_dp == expected_quantiles
), "unexpected weighted quantiles calculated"