def test_descriptive(self):
test('Descriptive Statistics')
series = [
to_series(binomial.generate(SIZE, n=10, p=0.2), name='Binomial'),
to_series(normal.generate(SIZE, mu=10, sigma=1), name='Normal'),
to_series(lognormal.generate(SIZE, mu=0, sigma=1),
name='Log-Normal'),
to_series(poisson.generate(SIZE, lam=10), name='Poisson')
]
df = concat_cols(series)
fig = descriptive.plot_multi_histogram(df,
share_x=False,
share_y=False)
fig.show()
fig = descriptive.plot_multi_density(df, share_x=False, share_y=False)
fig.show()
fig = descriptive.plot_histogram(get_col(df))
fig.show()
def test_poisson(self):
test(poisson.POISSON_NAME.title(), 'Distribution')
lam = 10
test(par(collist(lam)))
a = poisson.generate(SIZE, lam=lam)
s = to_series(a)
dist_a, dist_s = poisson.Poisson(series=a), poisson.Poisson(series=s)
test(dist_a, dist_s)
self.assert_dist(dist_a, dist_s)
interval = stats.poisson.interval(DEFAULT_CONFIDENCE_LEVEL, mu=lam)
test('- Real confidence interval:', interval)
def test_binomial(self):
test(binomial.BINOMIAL_NAME.title(), 'Distribution')
n = 10
p = 0.2
test(par(collist(n, p)))
a = binomial.generate(SIZE, n=n, p=p)
s = to_series(a)
dist_a, dist_s = binomial.Binomial(series=a), binomial.Binomial(
series=s)
test(dist_a, dist_s)
self.assert_dist(dist_a, dist_s)
interval = stats.binom.interval(DEFAULT_CONFIDENCE_LEVEL, n=n, p=p)
test('- Real confidence interval:', interval)
def test_mixture(self):
test('Fit a Gaussian mixture with one component')
data = np.random.randn(SIZE, 2)
data[:, 1] = exp(1 + data[:, 1])
data[:, 0], data[:, 1] = rotate(data[:, 0], data[:, 1], 1)
model = create_gaussian_mixture(data, component_count=1)
fig = plot_mixture(data, model, title='Gaussian Mixture With One Component')
fig.show()
test('Fit a Gaussian mixture with five components')
data = np.array([[0, -0.1], [1.7, 0.4]])
data = np.r_[np.dot(np.random.randn(SIZE, 2), data),
0.7 * np.random.randn(SIZE, 2) + np.array([-6, 3])]
model = create_gaussian_mixture(data, component_count=5)
fig = plot_mixture(data, model, title='Gaussian Mixture With Five Components')
fig.show()
test('Fit a Dirichlet process Gaussian mixture with five components')
data = np.array([[0, -0.1], [1.7, 0.4]])
data = np.r_[np.dot(np.random.randn(SIZE, 2), data),
0.7 * np.random.randn(SIZE, 2) + np.array([-6, 3])]
model = create_bayesian_gaussian_mixture(data, component_count=5)
fig = plot_mixture(data, model,
title='Dirichlet Process Gaussian Mixture With Five Components')
fig.show()
def test_normal_kde(self):
test(normal.NORMAL_KDE_NAME.title(), 'Distribution')
mu = 100
sigma = 10
test(par(collist(mu, sigma)))
a = normal.generate(SIZE, mu=mu, sigma=sigma)
s = to_series(a)
dist_a, dist_s = normal.NormalKDE(series=a), normal.NormalKDE(series=s)
test(dist_a, dist_s)
self.assert_dist(dist_a, dist_s)
interval = stats.norm.interval(DEFAULT_CONFIDENCE_LEVEL,
loc=mu,
scale=sigma)
test('- Real confidence interval:', interval)
def test_log_normal(self):
test(lognormal.LOG_NORMAL_NAME.title(), 'Distribution')
mu = 10
sigma = 1
test(par(collist(mu, sigma)))
a = lognormal.generate(SIZE, mu=mu, sigma=sigma)
s = to_series(a)
dist_a, dist_s = lognormal.LogNormal(series=a), lognormal.LogNormal(
series=s)
test(dist_a, dist_s)
self.assert_dist(dist_a, dist_s)
interval = stats.lognorm.interval(DEFAULT_CONFIDENCE_LEVEL,
s=sigma,
scale=exp(mu))
test('- Real confidence interval:', interval)
def assert_dist(self, dist_a, dist_s):
mean_a, mean_s = dist_a.mean(), dist_s.mean()
test('- Mean (with arrays):', mean_a)
test('- Mean (with series):', mean_s)
self.assert_equals(mean_a, mean_s)
std_a, std_s = dist_a.pdf(0), dist_s.pdf(0)
test('- Standard deviation (with arrays):', std_a)
test('- Standard deviation (with series):', std_s)
self.assert_equals(std_a, std_s)
entropy_a, entropy_s = dist_a.entropy(), dist_s.entropy()
test('- Entropy (with arrays):', entropy_a)
test('- Entropy (with series):', entropy_s)
self.assert_equals(entropy_a, entropy_s)
pdf_a, pdf_s = dist_a.pdf(1), dist_s.pdf(1)
test('- PDF (with arrays):', pdf_a)
test('- PDF (with series):', pdf_s)
self.assert_equals(pdf_a, pdf_s)
cdf_a, cdf_s = dist_a.cdf(0), dist_s.cdf(0)
test('- CDF (with arrays):', cdf_a)
test('- CDF (with series):', cdf_s)
self.assert_equals(cdf_a, cdf_s)
inv_cdf_a, inv_cdf_s = dist_a.inv_cdf(0.5), dist_s.inv_cdf(0.5)
test('- Inverse CDF (with arrays):', inv_cdf_a)
test('- Inverse CDF (with series):', inv_cdf_s)
self.assert_equals(inv_cdf_a, inv_cdf_s)
margin_a, margin_s = dist_a.margin(), dist_s.margin()
test('- Margin (with arrays):', margin_a)
test('- Margin (with series):', margin_s)
self.assert_equals(margin_a, margin_s)
interval_a, interval_s = dist_a.interval(mean=True), dist_s.interval(
mean=True)
test('- Mean confidence interval (with arrays):', interval_a)
test('- Mean confidence interval (with series):', interval_s)
self.assert_equals(interval_a, interval_s)
interval_a, interval_s = dist_a.interval(), dist_s.interval()
test('- Confidence interval (with arrays):', interval_a)
test('- Confidence interval (with series):', interval_s)
self.assert_equals(interval_a, interval_s)