"loc=outside" causes significance annotations to be cut out of the plot

[saskra]

Example:

Code:

sns.set_style('ticks')
ax = sns.violinplot(data=df_jpi, inner=None, color=".8")
sns.stripplot(data=df_jpi)
# ax.set(ylim=(0, 1))
test_results = add_stat_annotation(ax, 
                                                      data=df_jpi, 
                                                      box_pairs=[("X", "Y"), ("X", "Z"), ("Y", "Z")], 
                                                      test='Mann-Whitney', 
                                                      text_format='star', 
                                                      loc='outside', 
                                                      verbose=2)
fig = ax.get_figure()
fig.savefig('violin_out.png', transparent=True)
fig.show()

This does not happen with "loc=inside".

[trevismd]

Could you provide a small dataset to explore this ?

[saskra]

It does not depend on the dataset, so you can just use random data like these:

df_jpi = pd.DataFrame(np.random.randint(0, 100, size=(100, 3)), columns=list('XYZ'))

[trevismd]

You can use

fig.tight_layout()

to resize the figure to the added content before calling plt.show().
This is done with the parameter bbox_inches='tight' when saving the image (such as in the example notebook), but for plt.show(), this is the function to call first.

Thanks for reporting this. I'll look into automatically performing this step after adding annotations outside the plotting area (in statannotations).

[saskra]

Thanks a lot!