Creating dataframe boxplot from dataframe with row and column multiindex

Question

I have the following Pandas data frame and I'm trying to create a boxplot of the "dur" value for both client and server organized by qdepth (qdepth on x-axis, duration on y-axis, with two variables client and server). It seems like I need to get client and serveras columns. I haven't been able to figure this out trying combinations ofunstackandreset_index`.

It would help of if you could post a sample of your data in a format that we can import directly in pandas using pd.read_clipboard() (test it on your machine). Also, maybe a sketch of the boxplot you want to get would go a long way. — Julien Marrec
– Julien Marrec, Commented Nov 13, 2016 at 9:37

Julien Marrec · Accepted Answer · 2016-11-14 10:13:15Z

4

Here's some dummy data I recreated since you didn't post yours aside from an image:

qdepth,mode,runid,dur
1,client,0x1b7bd6ef955979b6e4c109b47690c862,7.0
1,client,0x45654ba030787e511a7f0f0be2db21d1,30.0
1,server,0xb760550f302d824630f930e3487b4444,19.0
1,server,0x7a044242aec034c44e01f1f339610916,95.0
2,client,0x51c88822b28dfa006bf38603d74f9911,15.0
2,client,0xd5a9028fddf9a400fd8513edbdc58de0,49.0
2,server,0x3943710e587e3932adda1cad8eaf2aeb,30.0
2,server,0xd67650fd984a48f2070de426e0a942b0,93.0

Load the data: df = pd.read_clipboard(sep=',', index_col=[0,1,2])

Option 1:

df.unstack(level=1).boxplot()

Option 2:

df.unstack(level=[0,1]).boxplot()

Option 3:

Using seaborn:

import seaborn as sns
sns.boxplot(x="qdepth", hue="mode", y="dur", data=df.reset_index(),)

Update:

To answer your comment, here's a very approximate way (could be used as a starting point) to recreate the seaborn option using only pandas and matplotlib:

fig, ax = plt.subplots(nrows=1,ncols=1, figsize=(12,6))
#bp = df.unstack(level=[0,1])['dur'].boxplot(ax=ax, return_type='dict')

bp = df.reset_index().boxplot(column='dur',by=['qdepth','mode'], ax=ax, return_type='dict')['dur']

# Now fill the boxes with desired colors
boxColors = ['darkkhaki', 'royalblue']
numBoxes = len(bp['boxes'])
for i in range(numBoxes):
    box = bp['boxes'][i]
    boxX = []
    boxY = []
    for j in range(5):
        boxX.append(box.get_xdata()[j])
        boxY.append(box.get_ydata()[j])
    boxCoords = list(zip(boxX, boxY))
    # Alternate between Dark Khaki and Royal Blue
    k = i % 2
    boxPolygon = mpl.patches.Polygon(boxCoords, facecolor=boxColors[k])
    ax.add_patch(boxPolygon)

plt.show()

edited Nov 14, 2016 at 10:13

answered Nov 13, 2016 at 10:11

Julien Marrec

11.9k5 gold badges51 silver badges67 bronze badges

Sign up to request clarification or add additional context in comments.

2 Comments

Noah Watkins Over a year ago

Thanks for the ansers. The Seaborn "hue" parameter seems to have been what I was looking for. It's not relevant for accepting this answer, but any way to achieve the groupings in vanilla Pandas without Seaborn?

Julien Marrec Over a year ago

Seaborn is litterally using vanilla matplotlib like pandas, so yes there is a way. You could look into Seaborn's source code (start here), and also this matplotlib example that shows how to color the boxes. (PS: an upvote would be nice in addition to accepting the answer)

Collectives™ on Stack Overflow

Creating dataframe boxplot from dataframe with row and column multiindex

1 Answer 1

2 Comments

Your Answer

Hot Network Questions

Collectives™ on Stack Overflow

1 Answer 1

2 Comments

Your Answer

Sign up or log in

Post as a guest

Related