def test_TA_odp(self):
model = apc.Model()
model.data_from_df(apc.loss_TA(), data_format='CL')
model.fit('od_poisson_response', 'AC')
model.simulate(repetitions=10)
model.simulate(repetitions=10, sigma2=10)
model.simulate(repetitions=10, od_poisson_dgp='neg_binomial')
def test_Belgian_gauss_rates(self):
model = apc.Model()
model.data_from_df(**apc.Belgian_lung_cancer())
model.fit('gaussian_rates', 'AC')
model.simulate(repetitions=10)
model.simulate(repetitions=10, fitted_values=model.fitted_values * 10)
model.simulate(repetitions=10, sigma2=10)
def test_VNJ_gln(self):
model = apc.Model()
model.data_from_df(apc.loss_VNJ(), data_format='CL')
model.fit('gen_log_normal_response', 'APC')
model.simulate(repetitions=10)
model.simulate(repetitions=10, fitted_values=model.fitted_values * 10)
model.simulate(repetitions=10, sigma2=10)
def test_BZ(self):
model = apc.Model()
model.data_from_df(apc.loss_BZ(), data_format='CL')
model.fit('od_poisson_response', 'AC')
model.forecast(method='n_poisson')
fc = model.forecast(attach_to_self=False)
self.assertEqual(fc['method'], 't_odp')
model.forecast([0.9])
def test_asbestos(self):
model = apc.Model()
model.data_from_df(apc.asbestos(), data_format='PA')
clone = model.clone()
for att in ('data_format', 'I', 'J', 'K', 'L', 'n', 'time_adjust'):
self.assertEqual(getattr(model, att), getattr(clone, att))
model.K = None
self.assertFalse(clone.K == model.K)
def test_Belgian_bin_dose_response(self):
data = apc.Belgian_lung_cancer()
dose = ((data['response'] / data['rate']) * 10**5).astype(int)
model = apc.Model()
model.data_from_df(data['response'], dose=dose, data_format='AP')
model.fit('binomial_dose_response', 'Ad')
model.simulate(repetitions=10)
model.simulate(repetitions=10, fitted_values=model.fitted_values * 10)
def test_Belgian(self):
model = apc.Model()
model.data_from_df(**apc.Belgian_lung_cancer())
model.plot_data_sums()
model.plot_data_sums(logy=True)
for simplify_ranges in ('start', 'mean', 'end', False):
model.plot_data_sums(simplify_ranges)
plt.close('all')
def test_TA(self):
model = apc.Model()
model.data_from_df(apc.loss_TA(), data_format='CL')
model.plot_data_sums()
model.plot_data_sums(logy=True)
for simplify_ranges in ('start', 'mean', 'end', False):
model.plot_data_sums(simplify_ranges)
plt.close('all')
def test_VNJ(self):
model = apc.Model()
model.data_from_df(apc.loss_VNJ(), data_format='CL')
model.fit('log_normal_response', 'AC')
sub_models = [model.sub_model(coh_from_to=(1,5)),
model.sub_model(coh_from_to=(6,10))]
f = apc.f_test(model, sub_models)
self.assertEqual(round(f['F_stat'], 3), 0.242)
self.assertEqual(round(f['p_value'], 3), 0.912)
def test_BZ(self):
model = apc.Model()
model.data_from_df(apc.loss_BZ(), data_format='CL')
model.fit('od_poisson_response', 'APC')
sub_models = [model.sub_model(per_from_to=(1977,1981)),
model.sub_model(per_from_to=(1982,1984)),
model.sub_model(per_from_to=(1985,1987))]
f = apc.f_test(model, sub_models)
self.assertEqual(round(f['F_stat'], 3), 1.855)
self.assertEqual(round(f['p_value'], 3), 0.133)
def test_TA_odp(self):
model = apc.Model()
model.data_from_df(apc.loss_TA(), data_format='CL')
model.fit(family='od_poisson_response', predictor='AC')
self.assertAlmostEqual(model.deviance, 1903014.004, 3)
self.assertTrue(
np.allclose(
model.parameters.sum().values,
np.array([10.99021854, 6.56495772, 0.37934664, 8.3525477])))
self.assertAlmostEqual(model.fitted_values.sum(), 34358090.000, 3)
def test_TA(self):
model = apc.Model()
model.data_from_df(apc.loss_TA(), data_format='CL')
model.fit('od_poisson_response', 'Ad')
model.forecast()
model.plot_forecast()
model.plot_forecast(ic=True)
model.plot_forecast('Age', aggregate=True)
model.plot_forecast('Cohort')
model.plot_forecast(from_to=(8, 13))
plt.close('all')
def test_asbestos(self):
model = apc.Model()
model.data_from_df(apc.asbestos())
self.assertEqual(model.data_format, 'PA')
self.assertEqual(model.I, 70)
self.assertEqual(model.J, 41)
self.assertEqual(model.K, 110)
self.assertEqual(model.L, 69)
self.assertEqual(model.n, 2870)
self.assertEqual(model.data_vector.sum()['response'], 32164)
def test_BZ_gln(self):
model = apc.Model()
model.data_from_df(apc.loss_BZ(), data_format='CL')
model.fit(family='gen_log_normal_response', predictor='APC')
self.assertAlmostEqual(model.deviance, -287.459, 3)
self.assertTrue(
np.allclose(
model.parameters.sum().values,
np.array([11.83624119, 1.26201587, 312.66337107, 12.1751463])))
self.assertAlmostEqual(model.fitted_values.sum(), 10214114.721, 3)
def test_TA(self):
model = apc.Model()
model.data_from_df(apc.loss_TA(), data_format='CL')
model.plot_data_heatmaps()
for simplify_ranges in ('start', 'mean', 'end', False):
model.plot_data_heatmaps(simplify_ranges)
plt.close('all')
for space in ('AC', 'AP', 'PA', 'PC', 'CA', 'CP'):
model.plot_data_heatmaps(space=space)
plt.close('all')
model.plot_data_heatmaps(vmin=20)
def test_Belgian(self):
model = apc.Model()
model.data_from_df(**apc.Belgian_lung_cancer())
model.plot_data_heatmaps()
for simplify_ranges in ('start', 'mean', 'end', False):
model.plot_data_heatmaps(simplify_ranges=simplify_ranges)
plt.close('all')
for space in ('AC', 'AP', 'PA', 'PC', 'CA', 'CP'):
model.plot_data_heatmaps(space=space)
plt.close('all')
model.plot_data_heatmaps(vmax=50)
def test_Belgian_ln_rates(self):
model = apc.Model()
model.data_from_df(**apc.Belgian_lung_cancer())
model.fit(family='log_normal_rates', predictor='APC')
self.assertAlmostEqual(model.deviance, -44.854, 3)
self.assertTrue(
np.allclose(
model.parameters.sum().values,
np.array([1.07781409, 7.31144396, 9.98467135, 15.2844013])))
self.assertAlmostEqual(model.fitted_values.sum(), 552.365, 3)
def test_Belgian_gauss_rates(self):
model = apc.Model()
model.data_from_df(**apc.Belgian_lung_cancer())
predictors = [
'APC', 'AP', 'AC', 'PC', 'Ad', 'Pd', 'Cd', 'A', 'P', 'C', 't',
'tA', 'tP', 'tC', '1'
]
for p in predictors:
model.fit('gaussian_rates', p)
model.identify('sum_sum')
model.identify('detrend')
def test_asbestos(self):
model = apc.Model()
model.data_from_df(apc.asbestos(), data_format='PA')
model.fit('poisson_response', 'AC')
model.simulate(repetitions=10)
model.simulate(repetitions=10, poisson_dgp='multinomial')
model.simulate(repetitions=10, seed=1)
self.assertTrue(
np.allclose(
model.draws,
model.simulate(repetitions=10, seed=1, attach_to_self=False)))
model.simulate(repetitions=10, fitted_values=model.fitted_values * 10)
def test_Belgian(self):
model = apc.Model()
model.data_from_df(**apc.Belgian_lung_cancer())
model.fit('log_normal_rates', 'APC')
for sr in ('start', 'mean', 'end', False):
model.plot_residuals(simplify_ranges=sr)
plt.close('all')
for sp in ('AC', 'AP', 'PA', 'PC', 'CA', 'CP'):
model.plot_data_heatmaps(space=sp)
plt.close('all')
for r in ('anscombe', 'deviance', 'pearson', 'response'):
model.plot_residuals(kind=r)
def test_Belgian_poisson_dose_response(self):
model = apc.Model()
model.data_from_df(**apc.Belgian_lung_cancer())
model.fit_table('poisson_dose_response', 'APC')
self.assertTrue(
np.allclose(
model.deviance_table.astype(float).sum().values,
np.array([
2.33052840e+04, 5.08000000e+02, 2.12588574e+00,
2.30019096e+04, 2.38000000e+02, 2.44351741e+00
])))
def test_asbestos(self):
model = apc.Model()
model.data_from_df(apc.asbestos(), data_format='PA')
model.fit('poisson_response', 'AC')
model.plot_parameters()
for style in ('detrend', 'sum_sum'):
model.plot_parameters(style)
plt.close('all')
for sr in ('start', 'mean', 'end', False):
model.plot_parameters(simplify_ranges=sr)
plt.close('all')
model.plot_parameters(around_coef=False)
def test_asbestos(self):
model = apc.Model()
model.data_from_df(apc.asbestos(), data_format='PA')
model.fit('poisson_response', 'AC')
model.forecast()
model.plot_forecast()
model.plot_forecast(ic=True)
model.plot_forecast('Age', aggregate=True)
model.plot_forecast('Cohort')
model.plot_forecast(from_to=(1986, 2032))
model.plot_forecast(figsize=(8, 6))
plt.close('all')
def test_TA(self):
model = apc.Model()
model.data_from_df(apc.loss_TA(), data_format='CL')
model.fit('od_poisson_response', 'AC')
for sr in ('start', 'mean', 'end', False):
model.plot_residuals(simplify_ranges=sr)
plt.close('all')
for sp in ('AC', 'AP', 'PA', 'PC', 'CA', 'CP'):
model.plot_data_heatmaps(space=sp)
plt.close('all')
for r in ('anscombe', 'deviance', 'pearson', 'response'):
model.plot_residuals(kind=r)
def test_TA(self):
model = apc.Model()
model.data_from_df(apc.loss_TA(), data_format='CL')
model.plot_data_within()
for simplify_ranges in ('start', 'mean', 'end', False):
model.plot_data_within(simplify_ranges=simplify_ranges)
plt.close('all')
for aggregate in ('mean', 'sum'):
model.plot_data_within(aggregate=aggregate)
plt.close('all')
model.plot_data_within(logy=True)
model.plot_data_within(n_groups='all')
def test_TA(self):
model = apc.Model()
model.data_from_df(apc.loss_TA(), data_format='CL')
model.fit('od_poisson_response', 'Ad')
model.plot_fit()
for style in ('detrend', 'sum_sum'):
model.plot_fit(style)
plt.close('all')
for sr in ('start', 'mean', 'end', False):
model.plot_fit(simplify_ranges=sr)
plt.close('all')
model.plot_fit(around_coef=False)
def test_Belgian(self):
model = apc.Model()
model.data_from_df(**apc.Belgian_lung_cancer())
model.plot_data_within()
for simplify_ranges in ('start', 'mean', 'end', False):
model.plot_data_within(simplify_ranges=simplify_ranges)
plt.close('all')
for aggregate in ('mean', 'sum'):
model.plot_data_within(aggregate=aggregate)
plt.close('all')
model.plot_data_within(logy=True)
model.plot_data_within(n_groups='all')
def test_Belgian(self):
model = apc.Model()
model.data_from_df(**apc.Belgian_lung_cancer())
model.fit('log_normal_rates', 'APC')
model.plot_fit()
for style in ('detrend', 'sum_sum'):
model.plot_fit(style)
plt.close('all')
for sr in ('start', 'mean', 'end', False):
model.plot_fit(simplify_ranges=sr)
plt.close('all')
model.plot_fit(around_coef=False)
def test_TA(self):
model = apc.Model()
model.data_from_df(apc.loss_TA())
self.assertEqual(model.data_format, 'CA')
self.assertEqual(model.I, 10)
self.assertEqual(model.J, 10)
self.assertEqual(model.K, 10)
self.assertEqual(model.L, 0)
self.assertEqual(model.n, 55)
self.assertEqual(model.time_adjust, 1)
self.assertAlmostEqual(model.data_vector.sum()['response'], 34358090.0,
3)
def test_BZ(self):
model = apc.Model()
model.data_from_df(apc.loss_BZ(), data_format='CL')
self.assertEqual(model.data_format, 'CL')
self.assertEqual(model.I, 11)
self.assertEqual(model.J, 11)
self.assertEqual(model.K, 11)
self.assertEqual(model.L, 0)
self.assertEqual(model.n, 66)
self.assertEqual(model.time_adjust, 0)
self.assertAlmostEqual(model.data_vector.sum()['response'], 10221194.0,
3)