def test_bootstrapmoment_skewness() -> None:
moment = BootstrapMoment(numsamples=10000, rng=1234)
values = np.arange(100, dtype=float)
moment.fill(values)
assert abs(moment.skewness().nominal - _skewness(values)) < 0.01
assert abs(np.average(moment.skewness().samples) -
_skewness(values)) < 10.0
def test_bootstrapmoment_array_shapes() -> None:
moment = BootstrapMoment(numsamples=3, rng=1234)
values = np.arange(100, dtype=float)
moment.fill(values)
assert (moment.mean().samples.shape == moment.std().samples.shape ==
moment.variance().samples.shape == moment.skewness().samples.shape
== (3, ))
def test_bootstrapmoment_standard_distributions(
with_weights: bool,
generator: Callable[[], np.ndarray],
mu: float,
sigma: float,
skewness: float,
) -> None:
moment = BootstrapMoment(numsamples=100, rng=1234)
values = generator()
if with_weights:
weight = np.random.default_rng(91011).uniform(size=values.shape)
moment.fill(values, weight=weight)
else:
moment.fill(values)
tolerance = 0.01
assert moment.mean().nominal == pytest.approx(mu,
rel=tolerance,
abs=tolerance)
assert np.average(moment.mean().samples) == pytest.approx(mu,
rel=tolerance,
abs=tolerance)
assert moment.std().nominal == pytest.approx(sigma,
rel=tolerance,
abs=tolerance)
assert np.average(moment.std().samples) == pytest.approx(sigma,
rel=tolerance,
abs=tolerance)
assert moment.skewness().nominal == pytest.approx(skewness,
rel=tolerance,
abs=tolerance)
assert np.average(moment.skewness().samples) == pytest.approx(
skewness, rel=tolerance, abs=tolerance)
def test_bootstrapmoment_std_deviation() -> None:
moment = BootstrapMoment(numsamples=10000, rng=1234)
values = np.arange(100, dtype=float)
moment.fill(values)
assert abs(moment.variance().nominal - np.var(values)) < 0.01
assert abs(np.average(moment.variance().samples) - np.var(values)) < 10.0
assert abs(moment.std().nominal - np.std(values)) < 0.01
assert abs(np.average(moment.std().samples) - np.std(values)) < 1.0
def test_bootstrapmoment_correlations_many_samples() -> None:
moment = BootstrapMoment(numsamples=10000, rng=1234)
values = np.arange(100, dtype=float)
moment.fill(values)
w = moment._sum_w.samples.view().flatten()
t1 = moment._sum_wt.samples.view().flatten()
t2 = moment._sum_wt2.samples.view().flatten()
t3 = moment._sum_wt3.samples.view().flatten()
cor = np.corrcoef([w, t1, t2, t3])
for row, column in itertools.combinations(range(4), 2):
assert cor[row, column] > 0.5
def test_bootstrapmoment_correlations(rng: Optional[int]) -> None:
moment = BootstrapMoment(numsamples=10000, rng=rng)
values = [1]
moment.fill(values)
w = moment._sum_w.samples.view().flatten()
t1 = moment._sum_wt.samples.view().flatten()
t2 = moment._sum_wt2.samples.view().flatten()
t3 = moment._sum_wt3.samples.view().flatten()
cor = np.corrcoef([w, t1, t2, t3])
for row, column in itertools.combinations(range(4), 2):
assert abs(cor[row, column] - 1.0) < 1e-6
def test_bootstrapmoment_add() -> None:
moment1 = BootstrapMoment(numsamples=1000, rng=1234)
moment2 = BootstrapMoment(numsamples=1000, rng=1234)
values1 = np.arange(50, dtype=float)
values2 = np.arange(50, 100, dtype=float)
moment1.fill(values1)
moment2.fill(values2)
moment = moment1 + moment2
values = np.concatenate([values1, values2])
assert abs(moment.mean().nominal - np.average(values)) < 0.001
assert abs(np.average(moment.mean().samples) - np.average(values)) < 1.0
assert abs(moment.variance().nominal - np.var(values)) < 0.01
assert abs(np.average(moment.variance().samples) - np.var(values)) < 10.0
assert abs(moment.std().nominal - np.std(values)) < 0.01
assert abs(np.average(moment.std().samples) - np.std(values)) < 1.0
assert abs(moment.skewness().nominal - _skewness(values)) < 0.01
assert abs(np.average(moment.skewness().samples) -
_skewness(values)) < 10.0
def test_pickle() -> None:
moment1 = BootstrapMoment(numsamples=1000, rng=1234)
values = np.arange(100, dtype=float)
moment1.fill(values)
moment2 = pickle.loads(pickle.dumps(moment1))
assert moment1 == moment2
def test_bootstrapmoment_numsamples() -> None:
numsamples = 1000
moment = BootstrapMoment(numsamples=numsamples, rng=1234)
values = np.arange(100, dtype=float)
moment.fill(values)
assert len(moment.mean().samples) == moment.numsamples == numsamples