Visualizing Distributions

There are several conventional ways to visualize distributions.

Histogram

Histogram is a layer that represents a histogram. It can be created by the add_hist method.

import numpy as np
from whitecanvas import new_canvas

np.random.seed(0)
data = np.random.normal(size=100)  # sample data

canvas = new_canvas("matplotlib")
canvas.add_hist(data, color="red")
canvas.show()

KDE

Kde is a layer that represents a kernel density estimation. It can be created by the add_kde method.

import numpy as np
from whitecanvas import new_canvas

np.random.seed(0)
data = np.random.normal(size=100)  # sample data

canvas = new_canvas("matplotlib")
canvas.add_kde(data, color="red")
canvas.show()

Rug

Rug is a layer that represents rug plot (or event plot). It can be created by the add_rug method. Rug plots tend to overlap with each other, so it is better at visualizing the individual data points rather than the distribution. Therefore, it is usually useful to overlay the rug plot on the histogram or the KDE.

import numpy as np
from whitecanvas import new_canvas

np.random.seed(0)
data = np.random.exponential(size=100)  # sample data

canvas = new_canvas("matplotlib")
canvas.add_hist(data, color="red")
canvas.add_rug(data)
canvas.show()

If you want to properly visualize the distribution only with the rug plot, you can use the color or length of the rug lines to represent the density of the data points. Rug layer has method color_by_density and scale_by_density to achieve this.

import numpy as np
from whitecanvas import new_canvas

np.random.seed(0)
data = np.random.normal(size=100)  # sample data

canvas = new_canvas("matplotlib")
canvas.add_rug(data).color_by_density(cmap="viridis")
canvas.show()

import numpy as np
from whitecanvas import new_canvas

np.random.seed(0)
data = np.random.normal(size=100)  # sample data

canvas = new_canvas("matplotlib")
canvas.add_rug(data).scale_by_density()
canvas.show()