min_subset_size expressed as a percentage / fraction

[MeHelmy]

Hi, Thank you for your support.

I tried to filter values with min_subset_size=0.1 when using show_percentages but it did not filter (will work if I set both show_percentages and show_count to True).

Another question (please let me know If I need to move it as separate one)

When using upset.add_stacked_bars and another method, ex: upset.add_catplot How I can control the legend location for each plot?

I used fig.legend(loc=(1, .7)) but it will work for one only, not both.

Thanks, Medhat

[jnothman]

Hi @MeHelmy, thanks for raising the issue. To help investigate, would you be able to provide complete, runnable code snippets that reproduce the issues?

[MeHelmy]

Thank you!

Here is an example:

import pandas as pd
import numpy as np
mydf = pd.DataFrame(np.random.randint(2,size=(1000, 4)), columns=list('ABCD'))
mydf['Country'] = pd.DataFrame(np.arange(1000).reshape(1000,1)).applymap(lambda x: np.random.choice(['Ge', 'Eg', 'Ni', 'UN']))

first=True
for i in mydf.columns.tolist()[:-1]:
    if first:
        first=False
        mydf = mydf.set_index(mydf[i] == 1)
    else:
        mydf = mydf.set_index(mydf[i] == 1,  append=True)

from upsetplot import UpSet
from matplotlib import cm
fig = plt.figure(figsize=(20, 10))
upset = UpSet(mydf,
              intersection_plot_elements=0,
              show_percentages=True,
              orientation='vertical',
              show_counts=False,
              min_subset_size=5,
              sort_by='degree',
             )  
upset.add_stacked_bars(by="Country", colors=cm.Pastel1,
                       title="Count by country", 
                        elements=20,
                      )
fig.legend(loc=(1, 1))
upset.plot()

That results in this plot:

So, I used min_subset_size=5 so I think I should not see 4.6% I tried to control the legend position looks like I am doing it wrong, lastly the annotation on the vertical bars are not clear can I change the annotation to be horizontal? (I know we can change the font by using mpl.rcParams['font.size'] = 10 or so, but this is not the case here).

Thanks, Medhat

[jnothman]

Are you saying that you want min_subset_size to also operate in percentages? Personally I find it very surprising behaviour if I had code

UpSet(mydf, min_subset_size=20)

and then added a parameter related to display

UpSet(mydf, min_subset_size=20, show_percentages=True)

and got an altogether different plot.

[jnothman]

Are you able to share code for your other issue with legends?

[MeHelmy]

Are you saying that you want min_subset_size to also operate in percentages? Personally I find it very surprising behaviour if I had code

UpSet(mydf, min_subset_size=20)

and then added a parameter related to display

UpSet(mydf, min_subset_size=20, show_percentages=True)

and got an altogether different plot.

I understand your point, let me put it this way, can I filter intersects to include only intersects representing more than 2% (just a random number here)?

[MeHelmy]

Are you able to share code for your other issue with legends?

It is in the same mentioned plot.

[jnothman]

We could support something like min_subset_frac=.2, or min_subset_size="20%", but we do not support these at present

[MeHelmy]

We could support something like min_subset_frac=.2, or min_subset_size="20%", but we do not support these at present

Sure! I was suggesting an idea :)

Best, Medhat

[jnothman]

I've renamed the issue in accordance with this.

[MeHelmy]

Thank you,

Do you have any suggestions for: fig.legend(loc=(1, 1)) and The annotation on the vertical bars, If I can change the annotation to be horizontal?

Thanks! Medhat

[jnothman]

I will admit legend placement is not within my expertise... Not sure what you're asking about annotations; label your picture?

[MeHelmy]

Thank you,

The annotation I mean upper left corner (red arrow).

[jnothman]

You want that text rotated? Currently you would have to get the axes returned by the plot() method, find the annotations and modify them...

[MeHelmy]

Thank you!

[jnothman]

I'm reopening this until we have a solution for min_subset_size expressed as a fraction

[jnothman]

I'm actually going to open a new issue so I don't need to keep reading this thread when I get occasion to work on this package .