Module ukat.utils.tests.test_arraystats
Functions
def calc_stats(x)-
Gold standard (gs) implementation of calculation of statistical measures.
Parameters
x:np.ndarray with 1 dimension
Notes
This is defined at the module level as it is used by both test Classes.
Classes
class TestArrayStats-
Expand source code
class TestArrayStats: # Good arrays array_4D = np.nan*np.ones((3, 3, 2, 2)) array_4D[:, :, 0, 0] = np.array([[4, 8, -1.2], [-4.8, 4, 2], [-3, 0, 5.0]]) array_4D[:, :, 0, 1] = np.array([[3, 6, -3.7], [-3.1, 2, 0], [-4, 4, 4.3]]) array_4D[:, :, 1, 0] = np.array([[2, 4, -2.2], [-1.7, 3, 3], [-2, 3, 3.2]]) array_4D[:, :, 1, 1] = np.array([[2, 2, -3.8], [-0.9, 4, 5], [-0, 1, 5.0]]) array_3D = np.nan*np.ones((3, 3, 2)) array_3D[:, :, 0] = np.array([[3, 6, -3.7], [-3.1, 2, 0], [-4, 4, 4.3]]) array_3D[:, :, 1] = np.array([[2, 4, -2.2], [-1.7, 3, 3], [-2, 3, 3.2]]) array_2D = np.array([[3, 6, -3.7], [-3.1, 2, 0], [-4, 4, 4.3]]) # Bad arrays array_5D = np.ones((2, 2, 2, 2, 2)) array_1D = np.array([1]) def test_4D_without_roi(self): # Calculate stats using ArrayStats stats = ast.ArrayStats(self.array_4D).calculate() # Calculate "gold standard" (gs) stats gs_4D = calc_stats(self.array_4D.flatten()) gs_3D_0 = calc_stats(self.array_4D[:, :, :, 0].flatten()) gs_3D_1 = calc_stats(self.array_4D[:, :, :, 1].flatten()) gs_2D_0 = calc_stats(self.array_4D[:, :, 0, 0].flatten()) gs_2D_1 = calc_stats(self.array_4D[:, :, 0, 1].flatten()) gs_2D_2 = calc_stats(self.array_4D[:, :, 1, 0].flatten()) gs_2D_3 = calc_stats(self.array_4D[:, :, 1, 1].flatten()) gs_list = [gs_4D, gs_3D_0, gs_3D_1, gs_2D_0, gs_2D_1, gs_2D_2, gs_2D_3] # Perform tests without ROI self.group_tests_4D(stats, gs_list) def test_4D_with_roi(self): # "Random" ROI roi_4D = np.ones((3, 3, 2, 2), dtype=bool) roi_4D[:, :, 0, 0] = np.array([[0, 0, 0], [1, 1, 1], [0, 0, 0]]) roi_4D[:, :, 0, 1] = np.array([[1, 0, 0], [1, 0, 0], [1, 0, 0]]) roi_4D[:, :, 1, 0] = np.array([[0, 1, 1], [0, 0, 0], [0, 0, 0]]) roi_4D[:, :, 1, 1] = np.array([[1, 1, 0], [1, 0, 0], [0, 0, 1]]) # Calculate stats using ArrayStats stats_roi = ast.ArrayStats(self.array_4D, roi=roi_4D).calculate() # Calculate "gold standard" (gs) stats gs_4D_roi = calc_stats(self.array_4D[roi_4D]) gs_3D_roi_0 = calc_stats(self.array_4D[:, :, :, 0][roi_4D[:, :, :, 0]]) gs_3D_roi_1 = calc_stats(self.array_4D[:, :, :, 1][roi_4D[:, :, :, 1]]) gs_2D_roi_0 = calc_stats(self.array_4D[:, :, 0, 0][roi_4D[:, :, 0, 0]]) gs_2D_roi_1 = calc_stats(self.array_4D[:, :, 0, 1][roi_4D[:, :, 0, 1]]) gs_2D_roi_2 = calc_stats(self.array_4D[:, :, 1, 0][roi_4D[:, :, 1, 0]]) gs_2D_roi_3 = calc_stats(self.array_4D[:, :, 1, 1][roi_4D[:, :, 1, 1]]) gs_list_roi = [gs_4D_roi, gs_3D_roi_0, gs_3D_roi_1, gs_2D_roi_0, gs_2D_roi_1, gs_2D_roi_2, gs_2D_roi_3] # Perform tests with ROI self.group_tests_4D(stats_roi, gs_list_roi) def group_tests_4D(self, stats, gs_list): """ Perform tests that compare outputs of ArrayStats.calculate() and the calc_stats() function above. Notes ----- This is used by tests with and without ROI so it was abstracted in its own function """ # Unpack list [gs_4D, gs_3D_0, gs_3D_1, gs_2D_0, gs_2D_1, gs_2D_2, gs_2D_3] = gs_list # Group gs stats to compare with output from ArrayStats gs_n_4D = gs_4D[0] gs_mean_4D = gs_4D[1] gs_median_4D = gs_4D[2] gs_minimum_4D = gs_4D[3] gs_maximum_4D = gs_4D[4] gs_std_4D = gs_4D[5] gs_cv_4D = gs_4D[6] gs_skewness_4D = gs_4D[7] gs_kurtosis_4D = gs_4D[8] gs_entropy_4D = gs_4D[9] gs_n_3D = np.array([gs_3D_0[0], gs_3D_1[0]]) gs_mean_3D = np.array([gs_3D_0[1], gs_3D_1[1]]) gs_median_3D = np.array([gs_3D_0[2], gs_3D_1[2]]) gs_minimum_3D = np.array([gs_3D_0[3], gs_3D_1[3]]) gs_maximum_3D = np.array([gs_3D_0[4], gs_3D_1[4]]) gs_std_3D = np.array([gs_3D_0[5], gs_3D_1[5]]) gs_cv_3D = np.array([gs_3D_0[6], gs_3D_1[6]]) gs_skewness_3D = np.array([gs_3D_0[7], gs_3D_1[7]]) gs_kurtosis_3D = np.array([gs_3D_0[8], gs_3D_1[8]]) gs_entropy_3D = np.array([gs_3D_0[9], gs_3D_1[9]]) gs_n_2D = np.array([[gs_2D_0[0], gs_2D_1[0]], [gs_2D_2[0], gs_2D_3[0]]]) gs_mean_2D = np.array([[gs_2D_0[1], gs_2D_1[1]], [gs_2D_2[1], gs_2D_3[1]]]) gs_median_2D = np.array([[gs_2D_0[2], gs_2D_1[2]], [gs_2D_2[2], gs_2D_3[2]]]) gs_minimum_2D = np.array([[gs_2D_0[3], gs_2D_1[3]], [gs_2D_2[3], gs_2D_3[3]]]) gs_maximum_2D = np.array([[gs_2D_0[4], gs_2D_1[4]], [gs_2D_2[4], gs_2D_3[4]]]) gs_std_2D = np.array([[gs_2D_0[5], gs_2D_1[5]], [gs_2D_2[5], gs_2D_3[5]]]) gs_cv_2D = np.array([[gs_2D_0[6], gs_2D_1[6]], [gs_2D_2[6], gs_2D_3[6]]]) gs_skewness_2D = np.array([[gs_2D_0[7], gs_2D_1[7]], [gs_2D_2[7], gs_2D_3[7]]]) gs_kurtosis_2D = np.array([[gs_2D_0[8], gs_2D_1[8]], [gs_2D_2[8], gs_2D_3[8]]]) gs_entropy_2D = np.array([[gs_2D_0[9], gs_2D_1[9]], [gs_2D_2[9], gs_2D_3[9]]]) # Tests npt.assert_allclose(stats["n"]["4D"], gs_n_4D, rtol=1e-6) npt.assert_allclose(stats["mean"]["4D"], gs_mean_4D, rtol=1e-6) npt.assert_allclose(stats["median"]["4D"], gs_median_4D, rtol=1e-6) npt.assert_allclose(stats["min"]["4D"], gs_minimum_4D, rtol=1e-6) npt.assert_allclose(stats["max"]["4D"], gs_maximum_4D, rtol=1e-6) npt.assert_allclose(stats["std"]["4D"], gs_std_4D, rtol=1e-6) npt.assert_allclose(stats["cv"]["4D"], gs_cv_4D, rtol=1e-6) npt.assert_allclose(stats["skewness"]["4D"], gs_skewness_4D, rtol=1e-6) npt.assert_allclose(stats["kurtosis"]["4D"], gs_kurtosis_4D, rtol=1e-6) npt.assert_allclose(stats["entropy"]["4D"], gs_entropy_4D, rtol=1e-6) npt.assert_allclose(stats["n"]["3D"], gs_n_3D, rtol=1e-6) npt.assert_allclose(stats["mean"]["3D"], gs_mean_3D, rtol=1e-6) npt.assert_allclose(stats["median"]["3D"], gs_median_3D, rtol=1e-6) npt.assert_allclose(stats["min"]["3D"], gs_minimum_3D, rtol=1e-6) npt.assert_allclose(stats["max"]["3D"], gs_maximum_3D, rtol=1e-6) npt.assert_allclose(stats["std"]["3D"], gs_std_3D, rtol=1e-6) npt.assert_allclose(stats["cv"]["3D"], gs_cv_3D, rtol=1e-6) npt.assert_allclose(stats["skewness"]["3D"], gs_skewness_3D, rtol=1e-6) npt.assert_allclose(stats["kurtosis"]["3D"], gs_kurtosis_3D, rtol=1e-6) npt.assert_allclose(stats["entropy"]["3D"], gs_entropy_3D, rtol=1e-6) npt.assert_allclose(stats["n"]["2D"], gs_n_2D, rtol=1e-6) npt.assert_allclose(stats["mean"]["2D"], gs_mean_2D, rtol=1e-6) npt.assert_allclose(stats["median"]["2D"], gs_median_2D, rtol=1e-6) npt.assert_allclose(stats["min"]["2D"], gs_minimum_2D, rtol=1e-6) npt.assert_allclose(stats["max"]["2D"], gs_maximum_2D, rtol=1e-6) npt.assert_allclose(stats["std"]["2D"], gs_std_2D, rtol=1e-6) npt.assert_allclose(stats["cv"]["2D"], gs_cv_2D, rtol=1e-6) npt.assert_allclose(stats["skewness"]["2D"], gs_skewness_2D, rtol=1e-6) npt.assert_allclose(stats["kurtosis"]["2D"], gs_kurtosis_2D, rtol=1e-6) npt.assert_allclose(stats["entropy"]["2D"], gs_entropy_2D, rtol=1e-6) def test_3D_without_roi(self): # Calculate stats using ArrayStats stats = ast.ArrayStats(self.array_3D).calculate() # Calculate "gold standard" (gs) stats gs_3D = calc_stats(self.array_3D.flatten()) gs_2D_0 = calc_stats(self.array_3D[:, :, 0].flatten()) gs_2D_1 = calc_stats(self.array_3D[:, :, 1].flatten()) # Group gs stats to compare with output from ArrayStats gs_n_3D = gs_3D[0] gs_mean_3D = gs_3D[1] gs_median_3D = gs_3D[2] gs_minimum_3D = gs_3D[3] gs_maximum_3D = gs_3D[4] gs_std_3D = gs_3D[5] gs_cv_3D = gs_3D[6] gs_skewness_3D = gs_3D[7] gs_kurtosis_3D = gs_3D[8] gs_entropy_3D = gs_3D[9] gs_n_2D = np.array([gs_2D_0[0], gs_2D_1[0]]) gs_mean_2D = np.array([gs_2D_0[1], gs_2D_1[1]]) gs_median_2D = np.array([gs_2D_0[2], gs_2D_1[2]]) gs_minimum_2D = np.array([gs_2D_0[3], gs_2D_1[3]]) gs_maximum_2D = np.array([gs_2D_0[4], gs_2D_1[4]]) gs_std_2D = np.array([gs_2D_0[5], gs_2D_1[5]]) gs_cv_2D = np.array([gs_2D_0[6], gs_2D_1[6]]) gs_skewness_2D = np.array([gs_2D_0[7], gs_2D_1[7]]) gs_kurtosis_2D = np.array([gs_2D_0[8], gs_2D_1[8]]) gs_entropy_2D = np.array([gs_2D_0[9], gs_2D_1[9]]) npt.assert_allclose(stats["n"]["3D"], gs_n_3D, rtol=1e-6) npt.assert_allclose(stats["mean"]["3D"], gs_mean_3D, rtol=1e-6) npt.assert_allclose(stats["median"]["3D"], gs_median_3D, rtol=1e-6) npt.assert_allclose(stats["min"]["3D"], gs_minimum_3D, rtol=1e-6) npt.assert_allclose(stats["max"]["3D"], gs_maximum_3D, rtol=1e-6) npt.assert_allclose(stats["std"]["3D"], gs_std_3D, rtol=1e-6) npt.assert_allclose(stats["cv"]["3D"], gs_cv_3D, rtol=1e-6) npt.assert_allclose(stats["skewness"]["3D"], gs_skewness_3D, rtol=1e-6) npt.assert_allclose(stats["kurtosis"]["3D"], gs_kurtosis_3D, rtol=1e-6) npt.assert_allclose(stats["entropy"]["3D"], gs_entropy_3D, rtol=1e-6) npt.assert_allclose(stats["n"]["2D"], gs_n_2D, rtol=1e-6) npt.assert_allclose(stats["mean"]["2D"], gs_mean_2D, rtol=1e-6) npt.assert_allclose(stats["median"]["2D"], gs_median_2D, rtol=1e-6) npt.assert_allclose(stats["min"]["2D"], gs_minimum_2D, rtol=1e-6) npt.assert_allclose(stats["max"]["2D"], gs_maximum_2D, rtol=1e-6) npt.assert_allclose(stats["std"]["2D"], gs_std_2D, rtol=1e-6) npt.assert_allclose(stats["cv"]["2D"], gs_cv_2D, rtol=1e-6) npt.assert_allclose(stats["skewness"]["2D"], gs_skewness_2D, rtol=1e-6) npt.assert_allclose(stats["kurtosis"]["2D"], gs_kurtosis_2D, rtol=1e-6) npt.assert_allclose(stats["entropy"]["2D"], gs_entropy_2D, rtol=1e-6) def test_2D_without_roi(self): # Calculate stats using ArrayStats stats = ast.ArrayStats(self.array_2D).calculate() # Calculate "gold standard" (gs) stats gs_2D = calc_stats(self.array_2D.flatten()) # Group gs stats to compare with output from ArrayStats gs_n_2D = gs_2D[0] gs_mean_2D = gs_2D[1] gs_median_2D = gs_2D[2] gs_minimum_2D = gs_2D[3] gs_maximum_2D = gs_2D[4] gs_std_2D = gs_2D[5] gs_cv_2D = gs_2D[6] gs_skewness_2D = gs_2D[7] gs_kurtosis_2D = gs_2D[8] gs_entropy_2D = gs_2D[9] npt.assert_allclose(stats["n"], gs_n_2D, rtol=1e-6) npt.assert_allclose(stats["mean"], gs_mean_2D, rtol=1e-6) npt.assert_allclose(stats["median"], gs_median_2D, rtol=1e-6) npt.assert_allclose(stats["min"], gs_minimum_2D, rtol=1e-6) npt.assert_allclose(stats["max"], gs_maximum_2D, rtol=1e-6) npt.assert_allclose(stats["std"], gs_std_2D, rtol=1e-6) npt.assert_allclose(stats["cv"], gs_cv_2D, rtol=1e-6) npt.assert_allclose(stats["skewness"], gs_skewness_2D, rtol=1e-6) npt.assert_allclose(stats["kurtosis"], gs_kurtosis_2D, rtol=1e-6) npt.assert_allclose(stats["entropy"], gs_entropy_2D, rtol=1e-6) def test_bad_array_dimensions(self): with pytest.raises(ValueError): ast.ArrayStats(self.array_1D) with pytest.raises(ValueError): ast.ArrayStats(self.array_5D) def test_bad_roi_dtype(self): array = np.ones((3, 3, 2)) roi = np.ones((3, 3, 2)) with pytest.raises(TypeError): ast.ArrayStats(array, roi) def test_roi_shape_mismatch(self): array = np.ones((3, 3, 2)) roi = np.ones((3, 3, 3), dtype=bool) with pytest.raises(ValueError): ast.ArrayStats(array, roi)Class variables
var array_1Dvar array_2Dvar array_3Dvar array_4Dvar array_5D
Methods
def group_tests_4D(self, stats, gs_list)-
Perform tests that compare outputs of ArrayStats.calculate() and the calc_stats() function above.
Notes
This is used by tests with and without ROI so it was abstracted in its own function
def test_2D_without_roi(self)def test_3D_without_roi(self)def test_4D_with_roi(self)def test_4D_without_roi(self)def test_bad_array_dimensions(self)def test_bad_roi_dtype(self)def test_roi_shape_mismatch(self)
class TestFlatStats-
Expand source code
class TestFlatStats: def test_array_with_one_element(self): # Note `gs`: gold standard # Int x1 = np.array([5]) gs_1 = calc_stats(x1) stats_1 = self.flatstats_to_array(ast.FlatStats(x1).calculate()) npt.assert_allclose(stats_1, gs_1, rtol=1e-6) # Float x2 = np.array([5.3]) gs_2 = calc_stats(x2) stats_2 = self.flatstats_to_array(ast.FlatStats(x2).calculate()) npt.assert_allclose(stats_2, gs_2, rtol=1e-6) # Negative number x3 = np.array([-5.3]) gs_3 = calc_stats(x3) stats_3 = self.flatstats_to_array(ast.FlatStats(x3).calculate()) npt.assert_allclose(stats_3, gs_3, rtol=1e-6) # Zero x4 = np.array([0]) gs_4 = calc_stats(x4) stats_4 = self.flatstats_to_array(ast.FlatStats(x4).calculate()) npt.assert_allclose(stats_4, gs_4, rtol=1e-6, equal_nan=True) # Nan x5 = np.array([np.nan]) gs_5 = calc_stats(x5) stats_5 = self.flatstats_to_array(ast.FlatStats(x5).calculate()) npt.assert_allclose(stats_5, gs_5, rtol=1e-6, equal_nan=True) def test_array_with_multiple_elements(self): # Ints x1 = np.array([5, 9, 7, 12, 19]) gs_1 = calc_stats(x1) stats_1 = self.flatstats_to_array(ast.FlatStats(x1).calculate()) npt.assert_allclose(stats_1, gs_1, rtol=1e-6) # Float x2 = np.array([5, 9, np.sqrt(7), 12, 19]) gs_2 = calc_stats(x2) stats_2 = self.flatstats_to_array(ast.FlatStats(x2).calculate()) npt.assert_allclose(stats_2, gs_2, rtol=1e-6) # Negative numbers x3 = np.array([5, 9, -np.sqrt(7), -12, 19]) gs_3 = calc_stats(x3) stats_3 = self.flatstats_to_array(ast.FlatStats(x3).calculate()) npt.assert_allclose(stats_3, gs_3, rtol=1e-6) # Zero x4 = np.array([5, 0, np.sqrt(7), -12, 19]) gs_4 = calc_stats(x4) stats_4 = self.flatstats_to_array(ast.FlatStats(x4).calculate()) npt.assert_allclose(stats_4, gs_4, rtol=1e-6, equal_nan=True) # Nan x5 = np.array([5, 0, np.sqrt(7), -12, np.nan]) gs_5 = calc_stats(x5) stats_5 = self.flatstats_to_array(ast.FlatStats(x5).calculate()) npt.assert_allclose(stats_5, gs_5, rtol=1e-6, equal_nan=True) def test_empty_array(self): expected_stats = np.array([0, # n np.nan, # mean np.nan, # median np.nan, # min np.nan, # max np.nan, # std np.nan, # cv np.nan, # skewness np.nan, # kurtosis np.nan]) # entropy x1 = np.array([]) stats_1 = self.flatstats_to_array(ast.FlatStats(x1).calculate()) npt.assert_allclose(stats_1, expected_stats, rtol=1e-6, equal_nan=True) def test_nonflat_array(self): x1 = None with pytest.raises(ValueError): ast.FlatStats(x1) x2 = np.array([[5, 9], [3, 4]]) with pytest.raises(ValueError): ast.FlatStats(x2) def flatstats_to_array(self, fs): n = fs.n mean = fs.mean median = fs.median minimum = fs.min maximum = fs.max std = fs.std cv = fs.cv skewness = fs.skewness kurtosis = fs.kurtosis entropy = fs.entropy return [n, mean, median, minimum, maximum, std, cv, skewness, kurtosis, entropy]Methods
def flatstats_to_array(self, fs)def test_array_with_multiple_elements(self)def test_array_with_one_element(self)def test_empty_array(self)def test_nonflat_array(self)