Cannot create a formatted pandas multindex barplot?

Question

I have a multiindex pandas dataframe that looks like this:

            cis-aconitate  cis-aconitate +1  ...  cis-aconitate +5  cis-aconitate +6
     GROUP                                   ...                                    
mean ADE         1.481639          0.696184  ...          0.193380          0.018597
     VEH         1.000000          1.000000  ...          1.000000          1.000000
std  ADE         0.307211          0.209418  ...          0.114939          0.020461
     VEH         0.573162          0.412895  ...          0.384928          0.752999

With indexing like this:

MultiIndex([('mean', 'ADE'),
            ('mean', 'VEH'),
            ( 'std', 'ADE'),
            ( 'std', 'VEH')],
           names=[None, 'GROUP'])

I am trying to create a barplot of the mean values on the y-axis. Using df.unstack() and then df.plot.bar() I can get this: But what I would ultimately like is to have the means only, with 'cis-aconitate, cis-aconitate +1 ...' across the x-axis, grouped into ADE and VEH (one color each) and then 'std' values used for the error bars. This is proving surprisingly tricky. Can anyone help? Thanks in advance!

danpl · Accepted Answer · 2021-10-12 17:39:57Z

1

To avoid doing a lot of transformations to the data frame and based on this example, the solution I found was to split the means from the errors using cross section.

import pandas as pd

# Example data
index = pd.MultiIndex.from_tuples([('mean', 'ADE'), ('mean', 'VEH'), ( 'std', 'ADE'), ('std', 'VEH')], names=[None, 'GROUP'])

df = pd.DataFrame({'cis-aconitate': [1.5, 1, 0.3, 0.2], 'cis-aconitate +1': [0.7, 1, 0.2, 0.4]}, index=index)
df

# Split dataframe into means and errors
means = df.xs('mean').transpose()
errors = df.xs('std').transpose()

means.plot.bar(yerr=errors)

answered Oct 12, 2021 at 17:39

danpl

3112 silver badges7 bronze badges

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

2 Comments

danpl Over a year ago

If you have access to the data that originated this data frame, seaborn can also be a nice alternative, as you'll probably only need to melt the data and have a 'group' column. It plots error bars as confidence intervals by default, but you can specify standard deviation.

jmorrr Over a year ago

This worked brilliantly! Thank you! Cross section with transpose was the key. I will definitely explore using seaborn with the long form data too. Will likely lead to more elegant code. Thanks very much for you help!

Collectives™ on Stack Overflow

Cannot create a formatted pandas multindex barplot?

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