ADD:
AIC statistic added
This commit is contained in:
@@ -232,11 +232,13 @@ class TestFitterPlotQQ:
|
||||
|
||||
def test_qq_hazen_returns_figure(self):
|
||||
import plotly.graph_objects as go
|
||||
|
||||
fig = self.f.plot_qq_plots(method="hazen")
|
||||
assert isinstance(fig, go.Figure)
|
||||
|
||||
def test_qq_filliben_returns_figure(self):
|
||||
import plotly.graph_objects as go
|
||||
|
||||
fig = self.f.plot_qq_plots(method="filliben")
|
||||
assert isinstance(fig, go.Figure)
|
||||
|
||||
@@ -257,11 +259,13 @@ class TestFitterHistogram:
|
||||
|
||||
def test_histogram_returns_figure(self):
|
||||
import plotly.graph_objects as go
|
||||
|
||||
fig = self.f.histogram_with_fits()
|
||||
assert isinstance(fig, go.Figure)
|
||||
|
||||
def test_histogram_seaborn_returns_figure(self):
|
||||
import matplotlib.pyplot as plt
|
||||
|
||||
fig = self.f.histogram_with_fits_seaborn()
|
||||
assert isinstance(fig, plt.Figure)
|
||||
plt.close("all")
|
||||
|
||||
112
etc/tests/test_statistics.py
Normal file
112
etc/tests/test_statistics.py
Normal file
@@ -0,0 +1,112 @@
|
||||
import numpy as np
|
||||
import pytest
|
||||
from scipy.stats import gamma, expon, norm
|
||||
import sys
|
||||
import os
|
||||
|
||||
sys.path.insert(0, os.path.join(os.path.dirname(__file__), ".."))
|
||||
|
||||
from tools.statistics import aic_statistic
|
||||
from fitting.fitter import Fitter
|
||||
|
||||
|
||||
RNG = np.random.default_rng(42)
|
||||
GAMMA_DATA = RNG.gamma(shape=2.0, scale=1.5, size=200)
|
||||
|
||||
|
||||
# ── aic_statistic unit tests ──────────────────────────────────────────────────
|
||||
|
||||
|
||||
class TestAicStatistic:
|
||||
def _fitted_dist(self, dist, data, **kwargs):
|
||||
"""Return a frozen distribution fitted to data."""
|
||||
params = dist.fit(data, **kwargs)
|
||||
return dist(*params)
|
||||
|
||||
def test_returns_float(self):
|
||||
frozen = self._fitted_dist(gamma, GAMMA_DATA, floc=0)
|
||||
result = aic_statistic(frozen, GAMMA_DATA, axis=0)
|
||||
assert isinstance(float(result), float)
|
||||
|
||||
def test_formula_correct(self):
|
||||
"""AIC = 2k - 2*log_likelihood."""
|
||||
frozen = self._fitted_dist(gamma, GAMMA_DATA, floc=0)
|
||||
k = len(frozen.args)
|
||||
log_likelihood = np.sum(frozen.logpdf(GAMMA_DATA), axis=0)
|
||||
expected = 2 * k - 2 * log_likelihood
|
||||
assert pytest.approx(aic_statistic(frozen, GAMMA_DATA, axis=0)) == expected
|
||||
|
||||
def test_penalises_more_parameters(self):
|
||||
"""gamma (3 params) should have higher AIC penalty term than expon (2 params)
|
||||
when both are fitted to the same data with identical log-likelihood contribution."""
|
||||
gamma_frozen = self._fitted_dist(gamma, GAMMA_DATA, floc=0)
|
||||
expon_frozen = self._fitted_dist(expon, GAMMA_DATA, floc=0)
|
||||
# penalty term alone: 2*k; gamma has more params so its penalty is larger
|
||||
assert 2 * len(gamma_frozen.args) > 2 * len(expon_frozen.args)
|
||||
|
||||
def test_better_fit_has_lower_aic(self):
|
||||
"""Gamma fitted to gamma data should have lower AIC than normal fitted to gamma data."""
|
||||
gamma_frozen = self._fitted_dist(gamma, GAMMA_DATA, floc=0)
|
||||
norm_frozen = self._fitted_dist(norm, GAMMA_DATA)
|
||||
aic_gamma = aic_statistic(gamma_frozen, GAMMA_DATA, axis=0)
|
||||
aic_norm = aic_statistic(norm_frozen, GAMMA_DATA, axis=0)
|
||||
assert aic_gamma < aic_norm
|
||||
|
||||
def test_works_with_axis_none(self):
|
||||
frozen = self._fitted_dist(gamma, GAMMA_DATA, floc=0)
|
||||
result = aic_statistic(frozen, GAMMA_DATA, axis=None)
|
||||
assert np.isfinite(result)
|
||||
|
||||
def test_result_is_finite(self):
|
||||
frozen = self._fitted_dist(gamma, GAMMA_DATA, floc=0)
|
||||
assert np.isfinite(aic_statistic(frozen, GAMMA_DATA, axis=0))
|
||||
|
||||
|
||||
# ── Integration: aic_statistic as callable in Fitter ─────────────────────────
|
||||
|
||||
|
||||
class TestAicStatisticInFitter:
|
||||
def test_fitter_accepts_aic_callable(self):
|
||||
f = Fitter([gamma], statistic_method=aic_statistic, gamma_params={"floc": 0})
|
||||
f.fit(GAMMA_DATA)
|
||||
f.validate(n_mc_samples=99)
|
||||
assert f["gamma"].test_result is not None
|
||||
|
||||
def test_fitter_aic_statistic_is_finite(self):
|
||||
f = Fitter([gamma], statistic_method=aic_statistic, gamma_params={"floc": 0})
|
||||
f.fit(GAMMA_DATA)
|
||||
f.validate(n_mc_samples=99)
|
||||
assert np.isfinite(f["gamma"].gof_statistic)
|
||||
|
||||
def test_fitter_aic_pvalue_in_range(self):
|
||||
f = Fitter([gamma], statistic_method=aic_statistic, gamma_params={"floc": 0})
|
||||
f.fit(GAMMA_DATA)
|
||||
f.validate(n_mc_samples=99)
|
||||
pval = f["gamma"].pvalue
|
||||
assert 0.0 <= pval <= 1.0
|
||||
|
||||
def test_fitter_aic_vs_ad_different_statistic_values(self):
|
||||
"""AIC and AD statistics should differ numerically."""
|
||||
f_aic = Fitter(
|
||||
[gamma], statistic_method=aic_statistic, gamma_params={"floc": 0}
|
||||
)
|
||||
f_ad = Fitter([gamma], statistic_method="ad", gamma_params={"floc": 0})
|
||||
f_aic.fit(GAMMA_DATA)
|
||||
f_ad.fit(GAMMA_DATA)
|
||||
f_aic.validate(n_mc_samples=99)
|
||||
f_ad.validate(n_mc_samples=99)
|
||||
assert f_aic["gamma"].gof_statistic != pytest.approx(
|
||||
f_ad["gamma"].gof_statistic
|
||||
)
|
||||
|
||||
def test_fitter_aic_multiple_distributions(self):
|
||||
f = Fitter(
|
||||
[gamma, expon],
|
||||
statistic_method=aic_statistic,
|
||||
gamma_params={"floc": 0},
|
||||
expon_params={"floc": 0},
|
||||
)
|
||||
f.fit(GAMMA_DATA)
|
||||
f.validate(n_mc_samples=99)
|
||||
assert f["gamma"].test_result is not None
|
||||
assert f["expon"].test_result is not None
|
||||
Reference in New Issue
Block a user