def test_compute_stats():
np.random.seed(1)
assert statistics.compute_stats([]) == (None, None)
assert statistics.compute_stats([15.0]) == (15.0, None)
for nsamples, true_mean in product([10, 50, 250], [0, 0.3, 0.6]):
samples = np.random.randn(nsamples) + true_mean
result, stats = statistics.compute_stats(samples)
assert np.allclose(stats['systematic'], 0)
assert np.allclose(stats['n'], len(samples))
assert np.allclose(stats['mean'], np.mean(samples))
assert np.allclose(stats['q_25'], np.percentile(samples, 25))
assert np.allclose(stats['q_75'], np.percentile(samples, 75))
assert np.allclose(stats['min'], np.min(samples))
assert np.allclose(stats['max'], np.max(samples))
assert np.allclose(stats['std'], np.std(samples, ddof=0))
assert np.allclose(result, np.median(samples))
ci = stats['ci_99']
assert ci[0] <= true_mean <= ci[1]
w = 12.0 * np.std(samples) / np.sqrt(len(samples))
assert ci[1] - ci[0] < w
err = statistics.get_err(result, stats)
iqr = np.percentile(samples, 75) - np.percentile(samples, 25)
assert np.allclose(err, iqr/2)
def test_compute_stats():
np = pytest.importorskip("numpy")
np.random.seed(1)
assert statistics.compute_stats([], 1) == (None, None)
assert statistics.compute_stats([15.0], 1) == (15.0, {
'ci_99_a': -math.inf,
'ci_99_b': math.inf,
'number': 1,
'q_25': 15.0,
'q_75': 15.0,
'repeat': 1
})
for nsamples, true_mean in product([10, 50, 250], [0, 0.3, 0.6]):
samples = np.random.randn(nsamples) + true_mean
result, stats = statistics.compute_stats(samples, 42)
assert stats['repeat'] == len(samples)
assert stats['number'] == 42
assert np.allclose(stats['q_25'], np.percentile(samples, 25))
assert np.allclose(stats['q_75'], np.percentile(samples, 75))
assert np.allclose(result, np.median(samples))
ci = stats['ci_99_a'], stats['ci_99_b']
assert ci[0] <= true_mean <= ci[1]
w = 12.0 * np.std(samples) / np.sqrt(len(samples))
assert ci[1] - ci[0] < w
err = statistics.get_err(result, stats)
iqr = np.percentile(samples, 75) - np.percentile(samples, 25)
assert np.allclose(err, iqr / 2)
def test_compute_stats():
np.random.seed(1)
assert statistics.compute_stats([], 1) == (None, None)
assert statistics.compute_stats([15.0], 1) == (15.0, {
'ci_99': [-inf, inf],
'max': 15.0,
'mean': 15.0,
'min': 15.0,
'number': 1,
'q_25': 15.0,
'q_75': 15.0,
'repeat': 1,
'std': 0.0
})
for nsamples, true_mean in product([10, 50, 250], [0, 0.3, 0.6]):
samples = np.random.randn(nsamples) + true_mean
result, stats = statistics.compute_stats(samples, 42)
assert stats['repeat'] == len(samples)
assert stats['number'] == 42
assert np.allclose(stats['mean'], np.mean(samples))
assert np.allclose(stats['q_25'], np.percentile(samples, 25))
assert np.allclose(stats['q_75'], np.percentile(samples, 75))
assert np.allclose(stats['min'], np.min(samples))
assert np.allclose(stats['max'], np.max(samples))
assert np.allclose(stats['std'], np.std(samples, ddof=0))
assert np.allclose(result, np.median(samples))
ci = stats['ci_99']
assert ci[0] <= true_mean <= ci[1]
w = 12.0 * np.std(samples) / np.sqrt(len(samples))
assert ci[1] - ci[0] < w
err = statistics.get_err(result, stats)
iqr = np.percentile(samples, 75) - np.percentile(samples, 25)
assert np.allclose(err, iqr / 2)