def test_basic_realizations_predictor(self):
"""
Test that the plugin returns a numpy float value. The ensemble
realizations are the predictor. The result indicates the minimum value
for the CRPS that was achieved by the minimisation.
"""
predictor = "realizations"
plugin = Plugin()
result = plugin.calculate_truncated_normal_crps(
self.initial_guess_for_realization,
self.forecast_predictor_data_realizations, self.truth_data,
self.forecast_variance_data, self.sqrt_pi, predictor)
self.assertIsInstance(result, np.float64)
self.assertAlmostEqual(result, 0.1670167)
def test_basic_mean_predictor_bad_value(self):
"""
Test that the plugin returns a numpy float64 value
and that the value matches the BAD_VALUE, when the appropriate
condition is found. The ensemble mean is the predictor. The initial
guess is specifically set to float32 precision for the purpose for
generating the BAD_VALUE for the unit test.
"""
initial_guess = np.array([1e65, 1e65, 1e65, 1e65], dtype=np.float32)
predictor = "mean"
plugin = Plugin()
result = plugin.calculate_truncated_normal_crps(
initial_guess, self.forecast_predictor_data, self.truth_data,
self.forecast_variance_data, self.sqrt_pi, predictor)
self.assertIsInstance(result, np.float64)
self.assertAlmostEqual(result, plugin.BAD_VALUE)
class Test_calculate_truncated_normal_crps(SetupTruncatedNormalInputs):
"""
Test minimising the crps for a truncated normal distribution.
Either the ensemble mean or the individual ensemble realizations are used
as the predictors.
"""
def setUp(self):
"""Set up plugin."""
super().setUp()
self.precision = 4
self.plugin = Plugin(tolerance=1e-4)
@ManageWarnings(
ignored_messages=["Collapsing a non-contiguous coordinate."])
def test_basic_mean_predictor(self):
"""
Test that the plugin returns a numpy float value. The ensemble mean
is the predictor. The result indicates the minimum value
for the CRPS that was achieved by the minimisation.
"""
predictor = "mean"
result = self.plugin.calculate_truncated_normal_crps(
self.initial_guess_for_mean,
self.forecast_predictor_data,
self.truth_data,
self.forecast_variance_data,
self.sqrt_pi,
predictor,
)
self.assertIsInstance(result, np.float64)
self.assertAlmostEqual(result, 0.2150, self.precision)
@ManageWarnings(
ignored_messages=["Collapsing a non-contiguous coordinate."])
def test_basic_realizations_predictor(self):
"""
Test that the plugin returns a numpy float value. The ensemble
realizations are the predictor. The result indicates the minimum value
for the CRPS that was achieved by the minimisation.
"""
predictor = "realizations"
result = self.plugin.calculate_truncated_normal_crps(
self.initial_guess_for_realization,
self.forecast_predictor_data_realizations,
self.truth_data,
self.forecast_variance_data,
self.sqrt_pi,
predictor,
)
self.assertIsInstance(result, np.float64)
self.assertAlmostEqual(result, 0.2150, self.precision)
@ManageWarnings(
ignored_messages=[
"Collapsing a non-contiguous coordinate.",
"invalid value encountered in",
],
warning_types=[UserWarning, RuntimeWarning],
)
def test_basic_mean_predictor_bad_value(self):
"""
Test that the plugin returns a numpy float64 value
and that the value matches the BAD_VALUE, when the appropriate
condition is found. The ensemble mean is the predictor. The initial
guess is specifically set to float32 precision for the purpose for
generating the BAD_VALUE for the unit test.
"""
initial_guess = np.array([1e65, 1e65, 1e65, 1e65], dtype=np.float32)
predictor = "mean"
result = self.plugin.calculate_truncated_normal_crps(
initial_guess,
self.forecast_predictor_data,
self.truth_data,
self.forecast_variance_data,
self.sqrt_pi,
predictor,
)
self.assertIsInstance(result, np.float64)
self.assertAlmostEqual(result, self.plugin.BAD_VALUE, self.precision)
