def mean_bias_removal(hindcast,
alignment,
cross_validate=True,
**metric_kwargs):
"""Calc and remove bias from py:class:`~climpred.classes.HindcastEnsemble`.
Args:
hindcast (HindcastEnsemble): hindcast.
alignment (str): which inits or verification times should be aligned?
- maximize/None: maximize the degrees of freedom by slicing ``hind`` and
``verif`` to a common time frame at each lead.
- same_inits: slice to a common init frame prior to computing
metric. This philosophy follows the thought that each lead should be
based on the same set of initializations.
- same_verif: slice to a common/consistent verification time frame prior
to computing metric. This philosophy follows the thought that each lead
should be based on the same set of verification dates.
cross_validate (bool): Use properly defined mean bias removal function. This
excludes the given initialization from the bias calculation. With False,
include the given initialization in the calculation, which is much faster
but yields similar skill with a large N of initializations.
Defaults to True.
Returns:
HindcastEnsemble: bias removed hindcast.
"""
def bias_func(a, b, **kwargs):
return a - b
bias_metric = Metric("bias", bias_func, True, False, 1)
bias = hindcast.verify(
metric=bias_metric,
comparison="e2o",
dim="init",
alignment=alignment,
**metric_kwargs,
).squeeze()
if cross_validate:
mean_bias_func = _mean_bias_removal_cross_validate
else:
mean_bias_func = _mean_bias_removal_quick
bias_removed_hind = mean_bias_func(hindcast._datasets["initialized"], bias,
"init")
bias_removed_hind = bias_removed_hind.squeeze()
if "dayofyear" in bias_removed_hind.coords:
del bias_removed_hind["dayofyear"]
hindcast_bias_removed = hindcast.copy()
hindcast_bias_removed._datasets["initialized"] = bias_removed_hind
return hindcast_bias_removed
from climpred.bootstrap import bootstrap_perfect_model
from climpred.comparisons import PM_COMPARISONS
from climpred.metrics import __ALL_METRICS__ as all_metrics, Metric, __pearson_r
from climpred.prediction import compute_hindcast, compute_perfect_model
def my_mse_function(forecast, reference, dim=None, **metric_kwargs):
return ((forecast - reference)**2).mean(dim)
my_mse = Metric(
name='mse',
function=my_mse_function,
positive=True,
probabilistic=False,
unit_power=2,
long_name='MSE',
aliases=['mSe', '<<<SE'],
)
ITERATIONS = 2
@pytest.mark.parametrize('comparison', PM_COMPARISONS)
def test_custom_metric_passed_to_compute(PM_da_initialized_1d,
PM_da_control_1d, comparison):
"""Test custom metric in compute_perfect_model."""
actual = compute_perfect_model(
PM_da_initialized_1d,
PM_da_control_1d,
from climpred.comparisons import PM_COMPARISONS
from climpred.metrics import __ALL_METRICS__ as all_metrics, Metric, __pearson_r
from climpred.stats import rm_poly
def my_mse_function(forecast, verif, dim=None, **metric_kwargs):
# function
return ((forecast - verif)**2).mean(dim)
my_mse = Metric(
name="mse",
function=my_mse_function,
positive=True,
probabilistic=False,
unit_power=2,
long_name="MSE",
aliases=["mSe", "<<<SE"],
)
ITERATIONS = 2
@pytest.mark.parametrize("comparison", PM_COMPARISONS)
def test_custom_metric_passed_to_verify(
perfectModelEnsemble_initialized_control, comparison):
"""Test custom metric in PerfectModelEnsemble.verify()."""
kwargs = dict(comparison=comparison, dim="init")
actual = perfectModelEnsemble_initialized_control.verify(metric=my_mse,
**kwargs)