def test_discrimination_sum(o, f_prob, dim, chunk_bool, input_type):
"""Test that the probabilities sum to 1"""
o, f_prob = modify_inputs(o, f_prob, input_type, chunk_bool)
if dim == []:
with pytest.raises(ValueError):
discrimination(o > 0.5, (f_prob > 0.5).mean("member"), dim=dim)
else:
disc = discrimination(o > 0.5, (f_prob > 0.5).mean("member"), dim=dim)
# test that input types equal output types
assign_type_input_output(disc, o)
if "Dataset" in input_type:
disc = disc[list(o.data_vars)[0]]
# dont understand the error message here, but it appeared
with suppress_warnings("invalid value encountered in true_divide"):
hist_event_sum = (
disc.sel(event=True).sum("forecast_probability", skipna=False).values
)
hist_no_event_sum = (
disc.sel(event=False).sum("forecast_probability", skipna=False).values
)
# Note, xarray's assert_allclose is already imported but won't compare to scalar
assert np.allclose(hist_event_sum[~np.isnan(hist_event_sum)], 1)
assert np.allclose(hist_no_event_sum[~np.isnan(hist_no_event_sum)], 1)
# test that returns chunks
assert_chunk(disc, chunk_bool)
def test_discrimination_perfect_values(o):
"""Test values for perfect forecast"""
f = xr.concat(10 * [o], dim="member")
disc = discrimination(o > 0.5, (f > 0.5).mean("member"))
assert np.allclose(disc.sel(event=True)[-1], 1)
assert np.allclose(disc.sel(event=True)[:-1], 0)
assert np.allclose(disc.sel(event=False)[0], 1)
assert np.allclose(disc.sel(event=False)[1:], 0)
def test_discrimination_accessor(o, f_prob, threshold, outer_bool):
actual = discrimination(o > threshold, (f_prob > threshold).mean("member"))
ds = xr.Dataset()
ds["o"] = o > threshold
ds["f_prob"] = (f_prob > threshold).mean("member")
if outer_bool:
ds = ds.drop_vars("f_prob")
expected = ds.xs.discrimination("o",
(f_prob > threshold).mean("member"))
else:
expected = ds.xs.discrimination("o", "f_prob")
assert_allclose(actual, expected)
def test_discrimination_sum(o, f_prob, dim, obj):
"""Test that the probabilities sum to 1"""
if "ds" in obj:
name = "var"
o = o.to_dataset(name=name)
f_prob = f_prob.to_dataset(name=name)
if "chunked" in obj:
o = o.chunk()
f_prob = f_prob.chunk()
if dim == []:
with pytest.raises(ValueError):
discrimination(o > 0.5, (f_prob > 0.5).mean("member"), dim=dim)
else:
disc = discrimination(o > 0.5, (f_prob > 0.5).mean("member"), dim=dim)
if "ds" in obj:
disc = disc[name]
hist_event_sum = (disc.sel(event=True).sum("forecast_probability",
skipna=False).values)
hist_no_event_sum = (disc.sel(event=False).sum("forecast_probability",
skipna=False).values)
# Note, xarray's assert_allclose is already imported but won't compare to scalar
assert np.allclose(hist_event_sum[~np.isnan(hist_event_sum)], 1)
assert np.allclose(hist_no_event_sum[~np.isnan(hist_no_event_sum)], 1)
def test_discrimination_dask(o_dask, f_prob_dask):
"""Test that discrimination returns dask array if provided dask array"""
disc = discrimination(o_dask > 0.5, (f_prob_dask > 0.5).mean("member"))
assert disc.chunks is not None