Merge pull request #27182 from i-jey/bihistogram_example

scottshambaugh · web-flow · commit 629baa83b2b2 · 2023-10-24T13:11:18.000-06:00
Add example for plotting a bihistogram
diff --git a/galleries/examples/statistics/histogram_bihistogram.py b/galleries/examples/statistics/histogram_bihistogram.py
@@ -0,0 +1,45 @@
+"""
+===========
+Bihistogram
+===========
+
+How to plot a bihistogram with Matplotlib.
+"""
+
+import matplotlib.pyplot as plt
+import numpy as np
+
+# Create a random number generator with a fixed seed for reproducibility
+rng = np.random.default_rng(19680801)
+
+# %%
+# Generate data and plot a bihistogram
+# ------------------------------------
+#
+# To generate a bihistogram we need two datasets (each being a vector of numbers).
+# We will plot both histograms using plt.hist() and set the weights of the second
+# one to be negative. We'll generate data below and plot the bihistogram.
+
+N_points = 10_000
+
+# Generate two normal distributions
+dataset1 = np.random.normal(0, 1, size=N_points)
+dataset2 = np.random.normal(1, 2, size=N_points)
+
+# Use a constant bin width to make the two histograms easier to compare visually
+bin_width = 0.25
+bins = np.arange(np.min([dataset1, dataset2]),
+                    np.max([dataset1, dataset2]) + bin_width, bin_width)
+
+fig, ax = plt.subplots()
+
+# Plot the first histogram
+ax.hist(dataset1, bins=bins, label="Dataset 1")
+
+# Plot the second histogram
+# (notice the negative weights, which flip the histogram upside down)
+ax.hist(dataset2, weights=-np.ones_like(dataset2), bins=bins, label="Dataset 2")
+ax.axhline(0, color="k")
+ax.legend()
+
+plt.show()
