4

How can we get legends for seaborn FacetGrid heatmaps? The .add_legend() method isn't working for me.

Using code from this previous question:

import pandas as pd
import numpy as np
import itertools
import seaborn as sns

print("seaborn version {}".format(sns.__version__))
# R expand.grid() function in Python
# https://stackoverflow.com/a/12131385/1135316
def expandgrid(*itrs):
   product = list(itertools.product(*itrs))
   return {'Var{}'.format(i+1):[x[i] for x in product] for i in range(len(itrs))}

methods=['method 1', 'method2', 'method 3', 'method 4']
times = range(0,100,10)
data = pd.DataFrame(expandgrid(methods, times, times))
data.columns = ['method', 'dtsi','rtsi']
data['nw_score'] = np.random.sample(data.shape[0])

def facet(data,color):
    data = data.pivot(index="dtsi", columns='rtsi', values='nw_score')
    g = sns.heatmap(data, cmap='Blues', cbar=False)

with sns.plotting_context(font_scale=5.5):
    g = sns.FacetGrid(data, col="method", col_wrap=2, size=3, aspect=1)
    g = g.map_dataframe(facet)
    g.add_legend()
    g.set_titles(col_template="{col_name}", fontweight='bold', fontsize=18)

enter image description here

Community
  • 1
  • 1
jmatsen
  • 103
  • 3
  • 7
  • Could you explain what sort of legend/legends you want? Legends are useful when you have multiple artists within a single set of axes. In your example there is only a single artist (heatmap) per subplot, and each subplot is uniquely titled with the corresponding column name, so it seems to me that a legend would be redundant in this case. – ali_m Jan 01 '16 at 01:57
  • You'll want to adapt the answer [here](http://stackoverflow.com/questions/28356359/one-colorbar-for-seaborn-heatmaps-in-subplot) – mwaskom Jan 01 '16 at 02:07
  • Thanks @mwaskom! I'm new to this and stoked to use it. – jmatsen Jan 01 '16 at 02:53

1 Answers1

11

What you want (in matplotlib lingo) is a colorbar, not a legend. In matplotlib, the former is used for continuous data, while the latter is used for categorical data. Colorbar support isn't built into FacetGrid, but it is not hard to expand your example code to add a colorbar:

import pandas as pd
import numpy as np
import itertools
import seaborn as sns

methods=['method 1', 'method2', 'method 3', 'method 4']
times = range(0, 100, 10)
data = pd.DataFrame(list(itertools.product(methods, times, times)))
data.columns = ['method', 'dtsi','rtsi']
data['nw_score'] = np.random.sample(data.shape[0])

def facet_heatmap(data, color, **kws):
    data = data.pivot(index="dtsi", columns='rtsi', values='nw_score')
    sns.heatmap(data, cmap='Blues', **kws)  # <-- Pass kwargs to heatmap

with sns.plotting_context(font_scale=5.5):
    g = sns.FacetGrid(data, col="method", col_wrap=2, size=3, aspect=1)

cbar_ax = g.fig.add_axes([.92, .3, .02, .4])  # <-- Create a colorbar axes

g = g.map_dataframe(facet_heatmap,
                    cbar_ax=cbar_ax,
                    vmin=0, vmax=1)  # <-- Specify the colorbar axes and limits

g.set_titles(col_template="{col_name}", fontweight='bold', fontsize=18)
g.fig.subplots_adjust(right=.9)  # <-- Add space so the colorbar doesn't overlap the plot

enter image description here

I've indicated the changes I made and the rationale for them as inline comments.

mwaskom
  • 46,693
  • 16
  • 125
  • 127