diff options
Diffstat (limited to 'numpy/lib/tests/test_histograms.py')
| -rw-r--r-- | numpy/lib/tests/test_histograms.py | 44 |
1 files changed, 32 insertions, 12 deletions
diff --git a/numpy/lib/tests/test_histograms.py b/numpy/lib/tests/test_histograms.py index 5b51763b2..67a94877d 100644 --- a/numpy/lib/tests/test_histograms.py +++ b/numpy/lib/tests/test_histograms.py @@ -431,7 +431,7 @@ class TestHistogramOptimBinNums(object): def test_empty(self): estimator_list = ['fd', 'scott', 'rice', 'sturges', - 'doane', 'sqrt', 'auto'] + 'doane', 'sqrt', 'auto', 'stone'] # check it can deal with empty data for estimator in estimator_list: a, b = histogram([], bins=estimator) @@ -447,11 +447,11 @@ class TestHistogramOptimBinNums(object): # Some basic sanity checking, with some fixed data. # Checking for the correct number of bins basic_test = {50: {'fd': 4, 'scott': 4, 'rice': 8, 'sturges': 7, - 'doane': 8, 'sqrt': 8, 'auto': 7}, + 'doane': 8, 'sqrt': 8, 'auto': 7, 'stone': 2}, 500: {'fd': 8, 'scott': 8, 'rice': 16, 'sturges': 10, - 'doane': 12, 'sqrt': 23, 'auto': 10}, + 'doane': 12, 'sqrt': 23, 'auto': 10, 'stone': 9}, 5000: {'fd': 17, 'scott': 17, 'rice': 35, 'sturges': 14, - 'doane': 17, 'sqrt': 71, 'auto': 17}} + 'doane': 17, 'sqrt': 71, 'auto': 17, 'stone': 20}} for testlen, expectedResults in basic_test.items(): # Create some sort of non uniform data to test with @@ -471,11 +471,11 @@ class TestHistogramOptimBinNums(object): precalculated. """ small_dat = {1: {'fd': 1, 'scott': 1, 'rice': 1, 'sturges': 1, - 'doane': 1, 'sqrt': 1}, + 'doane': 1, 'sqrt': 1, 'stone': 1}, 2: {'fd': 2, 'scott': 1, 'rice': 3, 'sturges': 2, - 'doane': 1, 'sqrt': 2}, + 'doane': 1, 'sqrt': 2, 'stone': 1}, 3: {'fd': 2, 'scott': 2, 'rice': 3, 'sturges': 3, - 'doane': 3, 'sqrt': 2}} + 'doane': 3, 'sqrt': 2, 'stone': 1}} for testlen, expectedResults in small_dat.items(): testdat = np.arange(testlen) @@ -499,7 +499,7 @@ class TestHistogramOptimBinNums(object): """ novar_dataset = np.ones(100) novar_resultdict = {'fd': 1, 'scott': 1, 'rice': 1, 'sturges': 1, - 'doane': 1, 'sqrt': 1, 'auto': 1} + 'doane': 1, 'sqrt': 1, 'auto': 1, 'stone': 1} for estimator, numbins in novar_resultdict.items(): a, b = np.histogram(novar_dataset, estimator) @@ -538,12 +538,32 @@ class TestHistogramOptimBinNums(object): xcenter = np.linspace(-10, 10, 50) outlier_dataset = np.hstack((np.linspace(-110, -100, 5), xcenter)) - outlier_resultdict = {'fd': 21, 'scott': 5, 'doane': 11} + outlier_resultdict = {'fd': 21, 'scott': 5, 'doane': 11, 'stone': 6} for estimator, numbins in outlier_resultdict.items(): a, b = np.histogram(outlier_dataset, estimator) assert_equal(len(a), numbins) + def test_scott_vs_ise(self): + """Verify that Scott's rule and the ISE based method converges for normally distributed data""" + + def nbins_ratio(seed, size): + rng = np.random.RandomState(seed) + x = rng.normal(loc=0, scale=2, size=size) + a, b = len(np.histogram(x, 'stone')[0]), len(np.histogram(x, 'scott')[0]) + return a / (a + b) + + ll = [[nbins_ratio(seed, size) for size in np.geomspace(start=10, stop=100, num=4).round().astype(int)] + for seed in range(256)] + + # the average difference between the two methods decreases as the dataset size increases. + assert_almost_equal(abs(np.mean(ll, axis=0) - 0.5), + [0.1065248, + 0.0968844, + 0.0331818, + 0.0178057], + decimal=3) + def test_simple_range(self): """ Straightforward testing with a mixture of linspace data (for @@ -555,11 +575,11 @@ class TestHistogramOptimBinNums(object): # Checking for the correct number of bins basic_test = { 50: {'fd': 8, 'scott': 8, 'rice': 15, - 'sturges': 14, 'auto': 14}, + 'sturges': 14, 'auto': 14, 'stone': 8}, 500: {'fd': 15, 'scott': 16, 'rice': 32, - 'sturges': 20, 'auto': 20}, + 'sturges': 20, 'auto': 20, 'stone': 80}, 5000: {'fd': 33, 'scott': 33, 'rice': 69, - 'sturges': 27, 'auto': 33} + 'sturges': 27, 'auto': 33, 'stone': 80} } for testlen, expectedResults in basic_test.items(): |