在 Pandas/matplotlib 中绘制一天中每小时的直方图

Question

I have a timeseries dataset of hourly granularity showing the price return for an asset during each hour of the day for a year.

I am trying to plot a distribution of returns for each hour of the day and thought there may be a way to group the returns by each hour and then to plot a histogram for each hour - so the output of the loop/function/method (ie the goal I am aiming to achieve) would be 24 distribution plot showing the distribution of returns for each hour over of the day across the entire time period.

My current dataframe is multi-indexed as Day , Hour (this may not be correct to accomplish my goal, and I can change it if required).

I am able to use groupby to get an hourly average of return over the timeframe ( df.groupby("Hour").mean() ) and thought I could use a similar method to plot my distributions.

Any suggestions of how to accomplish this goal would be appreciated

Answer 1

Example with very simple data (just 9-10 and 10-11 AM for 3 days):

import pandas as pd
import matplotlib.pyplot as plt

arrays = [[1, 1, 2, 2, 3, 3], [9, 10, 9, 10, 9, 10]]
ind = pd.MultiIndex.from_arrays(arrays, names=['Day', 'Hour'])
df = pd.DataFrame({'Price': [34, 35, 37, 31, 30, 29]}, index=ind)

for hour, group in df.groupby("Hour"):
    group.plot(kind='hist', bins=10, title=f'{hour}:00 - {hour+1}:00')
    plt.show()

---

Yes, groupby already precisely does what you want -- binning a DataFrame into groups. You can then easily go over those groups and plot each one individually. (In fact, for me calling .mean() on the list of groups is the less intuitive thing to do.). This is the output for printing each groupby group:

          Price
Day Hour       
1   9        34
    10       35
2   9        37
    10       31
3   9        30
    10       29

(9,           Price
Day Hour       
1   9        34
2   9        37
3   9        30)
(10,           Price
Day Hour       
1   10       35
2   10       31
3   10       29)