Add a category without data in it to a plot in seaborn - matplotlib

Question

I am making plotting some data as a catplot like this:
ax = sns.catplot(x='Kind', y='VAF', hue='Sample', jitter=True, data=df, legend=False)
The trouble is that some of the categories of 'VAF' contain no data, and the corresponding label is not added to the plot. Is there a way to retain the label but just not plot any points for it?
Here is a reproducible example to help explain:
x=pd.DataFrame({'Data':[1,3,4,6,3,2],'Number':['One','One','One','One','Three','Three']})
plt.figure()
ax = sns.catplot(x='Number', y='Data', jitter=True, data=x)
In this plot you can see that on the x-axis, samples One and Three are displayed. But imagine that there is also a sample Two that just had no data points in it. How can I display One, Two, and Three on the x-axis?

Answer 1

Order parameter
Of course one would need to know which categories are expected. Given a list of expected categories, one can use the order parameter to supply the expected categories.
import pandas as pd
import seaborn as sns
import matplotlib.pyplot as plt
df = pd.DataFrame({'Data':[1,3,4,6,3,2],
'Number':['One','One','One','One','Three','Three']})
exp_cats = ["One", "Two", "Three"]
ax = sns.stripplot(x='Number', y='Data', jitter=True, data=df, order=exp_cats)
plt.show()
Alternatives
The above works with matplotlib 2.2.3, but not with 3.0. It works again with the current development version (hence 3.1). For the moment, there are the following alternatives:
A. Looping over categories
Given a list of expected categories, one can just loop over them and plot a scatter of each category.
import numpy as np
import pandas as pd
import matplotlib.pyplot as plt
df = pd.DataFrame({'Data':[1,3,4,6,3,2],
'Number':['One','One','One','One','Three','Three']})
exp_cats = ["One", "Two", "Three"]
for i, cat in enumerate(exp_cats):
cdf = df[df["Number"] == cat]
x = np.zeros(len(cdf))+i+.2*(np.random.rand(len(cdf))-0.5)
plt.scatter(x, cdf["Data"].values)
plt.xticks(range(len(exp_cats)), exp_cats)
plt.show()
B. Map categories to numbers.
You can map the expected categories to numbers and plot numbers instead of categories.
import numpy as np
import pandas as pd
import matplotlib.pyplot as plt
df = pd.DataFrame({'Data':[1,3,4,6,3,2],
'Number':['One','One','One','One','Three','Three']})
exp_cats = ["One", "Two", "Three"]
df["IntNumber"] = df["Number"].map(dict(zip(exp_cats, range(len(exp_cats)))))
plt.scatter(df["IntNumber"] + .2*(np.random.rand(len(df))-0.5), df["Data"].values,
c = df["IntNumber"].values.astype(int))
plt.xticks(range(len(exp_cats)), exp_cats)
plt.show()
C. Appending missing categories to the dataframe
Finally you may append nan values to the dataframe to make sure each expected category appears in it.
import numpy as np
import pandas as pd
import seaborn as sns
import matplotlib.pyplot as plt
df = pd.DataFrame({'Data':[1,3,4,6,3,2],
'Number':['One','One','One','One','Three','Three']})
exp_cats = ["One", "Two", "Three"]
dfa = df.append(pd.DataFrame({'Data':[np.nan]*len(exp_cats), 'Number':exp_cats}))
ax = sns.stripplot(x='Number', y='Data', jitter=True, data=dfa, order=exp_cats)
plt.show()