Seaborn FacetGrid - Place single colorbar after last subplot

Question

I'm trying to add a colorbar to a grid of 3 seaborn plots. I am able to add the colorbar to the 3 individual plots, or squeeze a colour bar next to the 3rd plot. I would like to have a single colorbar after the 3rd plot, without changing the size of the last plot.

I got lots of good ideas from this answer, but couldn't solve my exact problem: SO Question/Answer

Here is my current code:

import seaborn as sns

def masked_vs_unmasked_facets(output_dir, merged_df, target_col, thresholds):
    # defining the maximal values, to make the plot square
    z_min = merged_df[['z_full', 'z_masked']].min(axis=0, skipna=True).min(skipna=True)
    z_max = merged_df[['z_full', 'z_masked']].max(axis=0, skipna=True).max(skipna=True)
    z_range_value = max(abs(z_min), abs(z_max))

    # Setting the column values to create the facet grid
    for i, val in enumerate(thresholds):
        merged_df.loc[merged_df.info_score_masked > val, 'PlotSet'] = i

    # Start the actual plots
    g = sns.FacetGrid(merged_df, col='PlotSet', size=8)

    def facet_scatter(x, y, c, **kwargs):
        kwargs.pop("color")
        plt.scatter(x, y, c=c, **kwargs)
        # plt.colorbar() for multiple colourbars

    vmin, vmax = 0, 1
    norm=plt.Normalize(vmin=vmin, vmax=vmax)

    g = (g.map(facet_scatter, 'z_full', 'z_masked', 'info_score_masked', norm=norm, cmap='viridis'))

    ax = g.axes[0]
    for ax in ax:
        ax.set_xlim([-z_range_value * 1.1, z_range_value * 1.1])
        ax.set_ylim([-z_range_value * 1.1, z_range_value * 1.1])
        ax.plot(ax.get_xlim(), ax.get_ylim(), ls="--", c=".3")

    plt.colorbar() # Single squashed colorbar
    plt.show()

masked_vs_unmasked_facets(output_dir, masking_results, 'info_score_masked', [0, 0.7, 0.9])

Single colorbar, but 3rd plot squashed Multiple colorbars, but crowded

Did you read https://stackoverflow.com/questions/13784201/matplotlib-2-subplots-1-colorbar — ImportanceOfBeingErnest, Apr 17 '18 at 10:31
Thanks - that has nearly got it. I can now see the colorbar in a separate axis, so the 3rd plot is no longer squashed. The problem is now that the colorbar is not being saved when I save the plot. Any tips on how I solve this? At the moment, I'm looking at using `subplots_adjust`. I'll add this to my answer. — geonaut, Apr 17 '18 at 13:08

geonaut · Accepted Answer · 2018-04-18T12:38:45.470

Following advice from @ImportanceOfBeingEarnest, I found that I needed to add another set of axes to the facetgrid, and then assign these axes to the colorbar. In order to get this extra element saved to the plot, I used the bbox_extra_artist kwarg for the tight bounding box. Another minor addition was a small clause to catch edge cases where one of my facets had no data. In this case, I appended an empty row with a single instance the category, so there was always at least 1 row for each category.

import seaborn as sns

def masked_vs_unmasked_facets(output_dir, merged_df, target_col, thresholds):
    z_min = merged_df[['z_full', 'z_masked']].min(axis=0, skipna=True).min(skipna=True)
    z_max = merged_df[['z_full', 'z_masked']].max(axis=0, skipna=True).max(skipna=True)
    z_range_value = max(abs(z_min), abs(z_max))

    for i, val in enumerate(thresholds):
        merged_df.loc[merged_df.info_score_masked > val, 'PlotSet'] = i
        # Catch instances where there are no values in category, to ensure all facets are drawn each time
        if i not in merged_df['PlotSet'].unique():
            dummy_row = pd.DataFrame(columns=merged_df.columns, data={'PlotSet': [i]})
            merged_df = merged_df.append(dummy_row)

    g = sns.FacetGrid(merged_df, col='PlotSet', size=8)

    def facet_scatter(x, y, c, **kwargs):
        kwargs.pop("color")
        plt.scatter(x, y, c=c, **kwargs)

    vmin, vmax = 0, 1
    norm=plt.Normalize(vmin=vmin, vmax=vmax)

    g = (g.map(facet_scatter, 'z_full', 'z_masked', 'info_score_masked', norm=norm, cmap='viridis'))

    titles = ["Correlation for all masked / unmasked z-score with {} above {}".format(target_col, threshold) for threshold in thresholds]

    axs = g.axes.flatten()
    for i, ax in enumerate(axs):
        ax.set_title(titles[i])
        ax.set_xlim([-z_range_value * 1.1, z_range_value * 1.1])
        ax.set_ylim([-z_range_value * 1.1, z_range_value * 1.1])
        ax.plot(ax.get_xlim(), ax.get_ylim(), ls="--", c=".3")


    cbar_ax = g.fig.add_axes([1.015,0.13, 0.015, 0.8])
    plt.colorbar(cax=cbar_ax)
    # extra_artists used here
    plt.savefig(os.path.join(output_dir, 'masked_vs_unmasked_scatter_final.png'), bbox_extra_artists=(cbar_ax,),  bbox_inches='tight')

masked_vs_unmasked_facets(output_dir, masking_results, 'info_score_masked', [0, 0.7, 0.9])

This gave me:

Seaborn FacetGrid - Place single colorbar after last subplot

1 Answers1

Linked