def test_different_shapes_fitted_and_transformed(self):
imputer = PerColumnImputer()
X = pd.DataFrame(index=list(range(10)))
X["a"] = np.ones(10)
imputer.fit(X)
X["b"] = np.ones(10)
self.assertRaises(ValueError, imputer.transform, X)
def test_standard_replacement_behavior(self):
imputer = PerColumnImputer()
data = [np.NINF, np.PINF, np.nan, 100.0, -100.0, 1.0, 1.0]
truth = [-100.0, 100.0, 1.0, 100.0, -100.0, 1.0, 1.0]
X = pd.DataFrame({"a": data})
true_X = pd.DataFrame({"a": truth})
imputer.fit(X)
selected_X = imputer.transform(X)
pdt.assert_frame_equal(selected_X, true_X)
def test_partial_preset_col_to_PINF_given(self):
data = [np.NINF, np.PINF, np.nan, 100.0, -100.0, 1.0, 1.0]
truth = [-100.0, 100.0, 1.0, 100.0, -100.0, 1.0, 1.0]
X = pd.DataFrame({"a": data})
true_X = pd.DataFrame({"a": truth})
col_to_max = {"a": 100}
imputer = PerColumnImputer(col_to_PINF_repl_preset=col_to_max)
imputer.fit(X)
selected_X = imputer.transform(X)
pdt.assert_frame_equal(selected_X, true_X)
def test_only_nans_and_infs(self):
imputer = PerColumnImputer()
X = pd.DataFrame(index=list(range(100)))
X["NaNs"] = np.nan * np.ones(100)
X["PINF"] = np.PINF * np.ones(100)
X["NINF"] = np.NINF * np.ones(100)
imputer.fit(X)
selected_X = imputer.transform(X)
self.assertTrue((selected_X.values == 0).all())
def test_only_subset_of_columns_given(self):
data = [np.NINF, np.PINF, np.nan, 100.0, -100.0, 1.0, 1.0]
truth_a = [-100.0, 100.0, 0.0, 100.0, -100.0, 1.0, 1.0]
truth_b = [-100.0, 100.0, 1.0, 100.0, -100.0, 1.0, 1.0]
X = pd.DataFrame({"a": data, "b": data})
true_X = pd.DataFrame({"a": truth_a, "b": truth_b})
col_to_median = {"a": 0}
imputer = PerColumnImputer(col_to_NAN_repl_preset=col_to_median)
imputer.fit(X)
selected_X = imputer.transform(X)
pdt.assert_frame_equal(selected_X, true_X)
def test_preset_has_higher_priority_than_fit(self):
data = [np.NINF, np.PINF, np.nan, 100.0, -100.0, 1.0, 1.0]
truth = [-100.0, 100.0, 0.0, 100.0, -100.0, 1.0, 1.0]
X = pd.DataFrame({"a": data})
true_X = pd.DataFrame({"a": truth})
col_to_median = {"a": 0}
imputer = PerColumnImputer(col_to_NAN_repl_preset=col_to_median)
imputer.fit(X)
selected_X = imputer.transform(X)
pdt.assert_frame_equal(selected_X, true_X)
def test_NAN_preset_contains_more_columns_than_dataframe_to_fit(self):
X = pd.DataFrame(index=list(range(10)))
X["a"] = np.ones(10)
col_to_median = {"a": 0, "b": 0}
imputer = PerColumnImputer(col_to_NAN_repl_preset=col_to_median)
self.assertRaises(ValueError, imputer.fit, X)
def test_with_numpy_array(self):
imputer = PerColumnImputer()
X = pd.DataFrame(index=list(range(100)))
X["NaNs"] = np.nan * np.ones(100)
X["PINF"] = np.PINF * np.ones(100)
X["NINF"] = np.NINF * np.ones(100)
X_numpy = X.values
imputer.fit(X)
selected_X = imputer.transform(X)
#re-initialize for new dicts
imputer = PerColumnImputer()
imputer.fit(X_numpy)
selected_X_numpy = imputer.transform(X_numpy)
npt.assert_array_equal(selected_X.values, selected_X_numpy.values)
self.assertTrue(selected_X_numpy.shape, (1, 100))
def test_only_parameters_of_last_fit_count(self):
data = [np.NINF, np.PINF, np.nan, 100.0, -100.0, 1.0, 1.0]
data_2 = [np.NINF, np.PINF, np.nan, 10.0, -10.0, 3.0, 3.0]
truth_a = [-10.0, 10.0, 3.0, 10.0, -10.0, 3.0, 3.0]
truth_b = [-10.0, 10.0, 3.0, 10.0, -10.0, 3.0, 3.0]
X = pd.DataFrame({"a": data, "b": data})
X_2 = pd.DataFrame({"a": data_2, "b": data_2})
true_X = pd.DataFrame({"a": truth_a, "b": truth_b})
imputer = PerColumnImputer()
imputer.fit(X)
imputer.fit(X_2)
selected_X = imputer.transform(X_2)
pdt.assert_frame_equal(selected_X, true_X)
def test_not_fitted(self):
imputer = PerColumnImputer()
X = pd.DataFrame()
self.assertRaises(NotFittedError, imputer.transform, X)