How to calculate percentages when total is not 100%?

Question

I'm very new to pandas and matplotlib.

I have applied an questionnaire, and in a certain question people were asked the social networks they use. Options were Facebook, Instagram, Twitter, and others. They could select more than an option.

I want to organize this data to plot a bar chart. I have used the following code:

listsocial = df["SocialNetworks"].str.split(', ', expand=True)

listsocial.head()

listsocial = 100*listsocial.stack().value_counts(normalize=True)

and then:

sns.set(font_scale=1.4)

ax = listsocial.plot(kind='bar', figsize=(15,7), color=('#009C3B'), grid=True)
ax.yaxis.set_major_formatter(mtick.PercentFormatter(decimals=False))
plt.xticks(rotation=80)
plt.suptitle('Most used social networks', fontsize=20)
plt.xlabel('Social network', fontsize=14, labelpad=20)
plt.ylabel('Respondents\n(%)', fontsize=14, labelpad=20)

plt.show()

However, the result does not take in account the fact people could answer more than an option, thus the total should not be not 100%. I want the chart to display data like: 70% uses Facebook, 60% uses Instagram, etc.

Thanks in advance.

Answer 1

Splitting and stacking is not the way to go in this case.

I would create separate columns for each social network of interest and assign True if it is included in the string (a sort of one-hot encoder)

social_networks = pd.DataFrame()
for sn in ['Facebook', 'Twitter', ...]:
    social_networks[sn] = df['SocialNetworks'].str.contains(sn)

Then you can get the percentage with

social_networks = social_networks.mean()

Answer 2

Instead of calling value_counts(normalize=True) you could divide by the number of rows:

from matplotlib import pyplot as plt
from matplotlib import ticker as mtick
import numpy as np
import pandas as pd
import seaborn as sns

networks = np.array(['facebook', 'twitter', 'instagram', 'other'])
socnetw = [", ".join(networks[np.random.randint(0, 2, 4, dtype=bool)]) for _ in range(100)]
df = pd.DataFrame({"SocialNetworks": socnetw})

listsocial = df["SocialNetworks"].str.split(', ', expand=True)
listsocial = 100 * listsocial.stack().value_counts() / len(listsocial)
listsocial = listsocial.iloc[:-1] # remove the last row (which contains the count for 'None')

sns.set(font_scale=1.4)

ax = listsocial.plot(kind='bar', figsize=(15, 7), color=('#009C3B'), grid=True)
ax.yaxis.set_major_formatter(mtick.PercentFormatter(decimals=False))
plt.xticks(rotation=80)
plt.suptitle('Most used social networks', fontsize=20)
plt.xlabel('Social network', fontsize=14, labelpad=20)
plt.ylabel('Respondents (%)', fontsize=14, labelpad=20)
plt.tight_layout()
plt.show()