Overplotting, Transparency, and Jitter

If we have a very large number of points to plot or our numeric variables are discrete-valued, then it is possible that using a scatterplot straightforwardly will not be informative. The visualization will suffer from overplotting, where the high amount of overlap in points makes it difficult to see the actual relationship between the plotted variables.

plt.scatter(data = df, x = 'disc_var1', y = 'disc_var2')

In the above plot, we can infer some kind of negative relationship between the two variables, but the degree of variability in the data and strength of relationship are fairly unclear. In cases like this, we may want to employ transparency and jitter to make the scatterplot more informative. Transparency can be added to a scatter call by adding the "alpha" parameter set to a value between 0 (fully transparent, not visible) and 1 (fully opaque).

plt.scatter(data = df, x = 'disc_var1', y = 'disc_var2', alpha = 1/5)

Where more points overlap, the darker the image will be. Here, we can now see that there is a moderate negative relationship between the two numeric variables. Values of 0 and 10 on the x-axis are much rarer than the central values.

As an alternative or companion to transparency, we can also add jitter to move the position of each point slightly from its true value. This is not a direct option in matplotlib's scatter function, but is a built-in option with seaborn's regplot function. x- and y- jitter can be added independently, and won't affect the fit of any regression function, if made:

sb.regplot(data = df, x = 'disc_var1', y = 'disc_var2', fit_reg = False,
           x_jitter = 0.2, y_jitter = 0.2, scatter_kws = {'alpha' : 1/3})

The jitter settings will cause each point to be plotted in a uniform ±0.2 range of their true values. Note that transparency has been changed to be a dictionary assigned to the "scatter_kws" parameter. This is necessary so that transparency is specifically associated with the scatter component of the regplotfunction.