def test_single_bound(sample_metric_frame):
"""Tests handling of invalid confidence interval."""
with pytest.raises(ValueError) as exc:
plot_metric_frame(
sample_metric_frame,
metrics=["Recall"],
conf_intervals=["Recall Bounds Single"],
)
assert str(exc.value) == _CONF_INTERVALS_MUST_BE_ARRAY
def test_invalid_metric_frame():
"""Tests handling of invalid metric frame."""
with pytest.raises(ValueError) as exc:
plot_metric_frame(
"not_metric_frame",
metrics=["Recall"],
conf_intervals=["Recall Bounds Flipped"],
)
assert str(exc.value) == _METRIC_FRAME_INVALID_ERROR
def test_flipped_bounds(sample_metric_frame):
"""Tests handling of flipped bounds for confidence intervals.
Flipped bounds are when upper_bound is lower than lower_bound.
"""
with pytest.raises(ValueError) as exc:
plot_metric_frame(
sample_metric_frame,
metrics=["Recall"],
conf_intervals=["Recall Bounds Flipped"],
)
assert str(exc.value) == _CONF_INTERVALS_FLIPPED_BOUNDS_ERROR
def test_single_ax_input(sample_metric_frame):
"""Tests plotting function works with single axis input."""
ax = matplotlib.pyplot.subplot()
ax = plot_metric_frame(
sample_metric_frame,
metrics=["Recall"],
conf_intervals=["Recall Bounds"],
ax=ax,
kind="bar",
colormap="Pastel1",
)
def test_multi_ax_input(sample_metric_frame):
"""Tests plotting function works with multiple axis input."""
fig, ax = matplotlib.pyplot.subplots(nrows=1, ncols=2)
ax[0].set_title("Recall Plot")
ax[0].set_ylabel("Recall")
ax[0].set_xlabel("Race")
ax[0].set_ylim((0, 1))
ax = plot_metric_frame(
sample_metric_frame,
metrics=["Recall", "Accuracy"],
conf_intervals=["Recall Bounds", "Accuracy Bounds"],
ax=ax,
kind="bar",
colormap="Pastel1",
)
def test_plotting_output(sample_metric_frame):
"""Tests for the correct output shape and output type."""
axs = plot_metric_frame(sample_metric_frame).flatten()
assert len(axs) == 3 # 3 is number of metrics that aren't arrays
assert isinstance(axs[0], matplotlib.axes.Axes)
"Accuracy": accuracy_score,
"Accuracy Bounds": accuracy_wilson,
}
metric_frame = MetricFrame(
metrics=metrics_dict,
y_true=y_test_true,
y_pred=y_test_pred,
sensitive_features=test_set_sex,
)
# %%
# Plotting
# ========
# Plot metrics without confidence intervals
# -----------------------------------------
plot_metric_frame(metric_frame, kind="point", metrics=["Recall", "Accuracy"])
# %%
# Plot metrics with confidence intervals (possibly asymmetric)
# ------------------------------------------------------------
plot_metric_frame(
metric_frame,
kind="bar",
metrics=["Recall", "Accuracy"],
conf_intervals=["Recall Bounds", "Accuracy Bounds"],
plot_ci_labels=True,
subplots=False,
)
plot_metric_frame(
metric_frame,