Visualizing skewed data

Question

The following lines of code gives the plot but that is skewed towards left.

sns.distplot(fraud_false['Amount'],
             hist=False, color='darkblue',
             kde_kws = {'shade': True, 'linewidth': 3})

sns.distplot(fraud_true['Amount'],
             hist=False, color='darkred',
             kde_kws = {'shade': True, 'linewidth': 3})

plt.title('Density Plot')
plt.xlabel('Amount')
plt.ylabel('Density')

Taking log of data makes the distribution normal. But how to take the original values of data in x axis of the plot instead of these new log values?

How do readers run your code without having the complete data? Since you are new to Stack Overflow, take some time in reading [How do I ask a good question?](https://stackoverflow.com/help/how-to-ask) — Sheldore, Feb 13 '20 at 14:31
Do I understand correctly that you want to keep the data log-transformed but on the x-axis of your distplot you want to have xticks that are not on log-transformed scale? — sim, Feb 13 '20 at 14:34
Also see [this post](https://stackoverflow.com/a/60132262/12046409) about how one could create log tick-labels for log-transformed data. — JohanC, Feb 13 '20 at 16:22

score 3 · Accepted Answer · answered Feb 13 '20 at 15:07

I hope I understand your question correctly:

You could either log-scale your entire axis. This will result in your bins in the distplot to also be log-scaled. Alternatively, you can log-transform your data and update your xticks to reflect the log-scale nature of your axis. Below shows this by example.

from scipy.stats import skewnorm
import numpy as np
from matplotlib import pyplot as plt
import seaborn as sbs

fig, ax = plt.subplots(1, 3, figsize=(10, 5))
data = skewnorm.rvs(size=1000, a=5)
sbs.distplot(data, ax=ax[0])
sbs.distplot(data, ax=ax[1])
ax[1].set_xscale("log")
sbs.distplot(np.log(data), ax=ax[2])
ax[2].set_xticklabels([round(d, 1) for d in np.exp(ax[2].get_xticks())]);

Visualizing skewed data

1 Answers1