Plots for categorical features show only points, not lines. This is a good default. However, for ordered categoricals (or to stress that a set of points belongs together), it would be nice to show optional lines. On the other hand, it could make sense to optionally hide the lines for numeric features.
Example
import numpy as np
import pandas as pd
import matplotlib.pyplot as plt
from sklearn.linear_model import LinearRegression
from sklearn.preprocessing import OneHotEncoder
from sklearn.compose import make_column_transformer
from sklearn.pipeline import make_pipeline
from model_diagnostics.calibration import plot_marginal
fig, axes = plt.subplots(1, 2, figsize=(10, 5))
n = 1000
x_cat_values = (list("abc"), [0, 1, 2])
for i, values in enumerate(x_cat_values):
rng = np.random.default_rng(0)
X = pd.DataFrame(
{
"x_cat": rng.choice(values, n),
"x_num": rng.standard_normal(n),
}
)
y = (X.x_cat == values[0]) + 0.5 * X.x_num + rng.standard_normal(n)
model = make_pipeline(
make_column_transformer(
(OneHotEncoder(drop="first", sparse_output=False), ["x_cat"]),
remainder="passthrough"
),
LinearRegression()
).fit(X, y=y)
plot_marginal(
y,
y_pred=model.predict(X),
X=X,
feature_name="x_cat",
predict_function=model.predict,
ax=axes[i],
)
Proposal
Add argument show_lines=None to plots. By default, it would be False for categoricals, and True otherwise.
Plots for categorical features show only points, not lines. This is a good default. However, for ordered categoricals (or to stress that a set of points belongs together), it would be nice to show optional lines. On the other hand, it could make sense to optionally hide the lines for numeric features.
Example
Proposal
Add argument
show_lines=None
to plots. By default, it would be False for categoricals, and True otherwise.