def test_disjoint_inits(small_initialized_da, small_verif_da, alignment):
"""Tests that alignment works with disjoint inits in the verification
data, i.e., non-continuous initializing to verify with."""
hind = small_initialized_da.drop_sel(init=1991)
verif = small_verif_da
actual = compute_hindcast(hind, verif, alignment=alignment, metric="mse")
assert actual.notnull().all()
# hindcast inits: [1990, 1992, 1993]
# verif times: [1990, 1991, 1992, 1993, 1994]
a = hind.rename({"init": "time"})
b = verif.sel(time=[1991, 1993, 1994])
a["time"] = b["time"]
expected = xs.mse(a, b, "time")
assert actual == expected
def test_disjoint_verif_time(small_initialized_da, small_verif_da, alignment):
"""Tests that alignment works with disjoint time in the verification
data, i.e., non-continuous time sampling to verify against."""
hind = small_initialized_da
verif = small_verif_da.drop_sel(time=1992)
actual = (HindcastEnsemble(hind).add_observations(verif).verify(
comparison="e2o", dim="init", alignment=alignment, metric="mse"))
assert actual.notnull().all()
# hindcast inits: [1990, 1991, 1992, 1993]
# verif times: [1990, 1991, 1993, 1994]
a = hind.sel(init=[1990, 1992, 1993]).rename({"init": "time"})
b = verif.sel(time=[1991, 1993, 1994])
a["time"] = b["time"]
expected = xs.mse(a, b, "time")
assert actual == expected
def _mse(forecast, reference, dim='svd', comparison=None):
"""
Calculate the Mean Sqaure Error (MSE).
.. math::
MSE = \\overline{(f - o)^{2}}
Range:
* perfect: 0
* min: 0
* max: ∞
See also:
* xskillscore.mse
"""
return mse(forecast, reference, dim=dim)
def mse(x, y, dim):
"""
Compute Mean Squared Error.
Parameters
----------
x : Dataset, DataArray, GroupBy, Variable, numpy/dask arrays or scalars
Mix of labeled and/or unlabeled arrays to which to apply the function.
y : Dataset, DataArray, GroupBy, Variable, numpy/dask arrays or scalars
Mix of labeled and/or unlabeled arrays to which to apply the function.
dim : str
The dimension to apply the correlation along.
Returns
-------
Mean Squared Error
Single value or tuple of Dataset, DataArray, Variable, dask.array.Array or
numpy.ndarray, the first type on that list to appear on an input.
"""
return xs.mse(x, y, dim)
r = xs.pearson_r(obs, fct, "nlocs") # 2-tailed p-value of Pearson's correlation coefficient #jkim r_p_value = xs.pearson_r_p_value(obs, fct, "nlocs") # Spearman's correlation coefficient rs = xs.spearman_r(obs, fct, "nlocs") # 2-tailed p-value associated with Spearman's correlation coefficient #jkim rs_p_value = xs.spearman_r_p_value(obs, fct, "nlocs") # Root Mean Squared Error rmse = xs.rmse(obs, fct, "nlocs") # Mean Squared Error mse = xs.mse(obs, fct, "nlocs") # Mean Absolute Error mae = xs.mae(obs, fct, "nlocs") # Median Absolute Error median_absolute_error = xs.median_absolute_error(obs, fct, "nlocs") # Mean Absolute Percentage Error mape = xs.mape(obs, fct, "nlocs") # Symmetric Mean Absolute Percentage Error #jkim smape = xs.smape(obs, fct, "nlocs") # You can also specify multiple axes for deterministic metrics: # Apply Pearson's correlation coefficient