As I hope you can see from the image, I have a bundle of functions of 'x'. For each input point, there is a distribution of these functions, thus average (AVG) and variance (VAR). I would like to plot histograms on top of the graph of the functions like shown on the right of the image using matplotlib if possible, but HOW?

CodePudding user response：

You can use scipy.stats.gausian_kde to calculate a kernel density estimation and draw those vertically.

Here is some example code using 20 curves:

from matplotlib import pyplot as plt
import numpy as np
from scipy.stats import gaussian_kde

N = 100
K = 20
ys = np.random.normal(0.1, 1, (N, K)).cumsum(axis=0)
ys -= ys.min(axis=0, keepdims=True)
x = np.arange(N)
plt.plot(x, ys, color='cornflowerblue', alpha=0.2)
plt.plot(x, ys.mean(axis=1), color='navy', lw=3, label='mean y')
for i in range(10, len(x), 20):
    yi = ys[i, :]
    ymin, ymax = yi.min(), yi.max()
    y_range = ymax - ymin
    ymin -= y_range * 0.1
    ymax  = y_range * 0.1
    kde = gaussian_kde(yi, bw_method=y_range / 100)
    plt.vlines(x[i], ymin, ymax, ls=':', color='black')
    y = np.linspace(ymin, ymax, 200)
    kde_y = kde(y)
    kde_y *= - 10 / kde(y).max()
    kde_y  = x[i]
    plt.plot(kde_y, y, color='black')
    plt.fill_betweenx(y, x[i], kde_y, color='black', alpha=.2)
plt.xticks(x[::10])
plt.show()