How to get pandas to plot on the same graph with the same y axis range

user_1473517 0
Viewed 1201

I am trying to plot multiple bar charts vertically on top of each other. There should be one labelled x axis (with the days of the week). The code I have so far is:

import pandas as pd
import matplotlib.pyplot as plt
import calendar

df = pd.read_csv("health.csv", header = None, names = ['Physical', 'Emotional'])
# Get Dayofweek index number (start with 6 for sunday) 6,0,1....
df['DayOfTheWeek'] = [(i+6) % 7  for i in range(len(df))]

# Get a map to translate to day of week
d = dict(zip(range(7),list(calendar.day_name)))
df['DayOfTheWeek'] = df['DayOfTheWeek'].map(d)

# Loop through the df (splitting week by week)
for i in range(int(round(len(df)/7))):
    plt.ylim([0,10])
    df.iloc[i*7:(i+1)*7].set_index('DayOfTheWeek').plot(kind='bar')
plt.show()

This has the following problems:

  1. For some reasons the first graph produced is blank.
  2. I would like subplots on the same graph separated vertically rather than lots of separate plots
  3. My dataframe has 39 rows but the method above doesn't plot the last 4 points at all.

The full input data is:

5,5
6,7
6,9
6,7
5,6
7,9
5,9
6,7
7,6
7,4
7,5
6,7
7,9
7,9
5,6
8,7
9,9
7,7
7,6
7,8
7,9
7,9
7,6
7,8
6,6
6,6
6,7
6,6
6,5
6,6
7,5
7,5
7,5
7,6
7,5
8,6
7,6
7,7
6,6
pandas matplotlib python
2 Answers

1. For some reasons the first graph produced is blank.

When you call plt.ylim(), it will "set the y-limits of the current axes.". It does this by calling plt.gca under the hood, which will "Get the current Axes instance (...), or create one.". Now, in the first iteration of your loop, no Axes exists, so it creates a new one. Then pandas.DataFrame.plot proceeds to create its own figure, ignoring the existing one. That's how you get an empty first plot.

The fix is simple: Swap the order of plt.ylim([0,10]) and the following line, or set it directly in .plot(kind='bar', ylim=(0, 10)).

2. I would like subplots on the same graph separated vertically rather than lots of separate plots

Perhaps plt.subplots() is what you're looking for?

n_weeks = 6  # See pt 3 for an elaboration on this
fig, axs = plt.subplots(n_weeks, 1, figsize=(5, 12), sharex=True)

# Record the names of the first 7 days in the dataset
weekdays = df.head(7)['DayOfTheWeek'].values
for weekno, ax in enumerate(axs):
    week = df.iloc[weekno*7:(weekno+1)*7]
    week = week.set_index('DayOfTheWeek')
    # The final week is incomplete and will mess up our plot unless
    # we force it to contain all the weekdays.
    week = week.loc[weekdays]
    week.plot(kind='bar', ylim=(0, 10), ax=ax, legend=False)
# Only draw legend in the final Axis
ax.legend()

# Force tight layout
fig.tight_layout()

3. My dataframe has 39 rows but the method above doesn't plot the last 4 points at all.

Try printing the ranges you select in your loop, and you should be able to spot the error. It is an off-by-one error :-)

Spoiler/solution below!

for i in range(int(round(len(df)/7))):
    print(df.iloc[i*7:(i+1)*7])

shows that you are only selecting complete weeks.

Note: In copying the data from the question, I apparently missed a row! There should be 39. The remarks still stand, though.

Let's inspect what happens! len(df) is 38, len(df) / 7 is 5.43, and round(len(df) / 7) is 5. You are rounding down to nearest complete week. Had your data contained one more day, it would round up to 6 as you expect. However, that is somewhat brittle behaviour; sometimes it rounds up, sometimes down, but you always want to see the last incomplete week. So rather than doing that, I'll introduce you to two nice features: the // operator, which is a floor division (always rounding down), and divmod, a built-in function that simultaneously does floor division and gives you the remainder.

My suggested solution uses divmod to count any incomplete weeks:

n_weeks, remaining_days = divmod(len(df), 7)
n_weeks += min(1, remaining_days)

for i in range(n_weeks):
    ...
user_32759670

You can do this by first setting up your figure layout, then passing an explicit axes object to the pandas plot method. I then conditionally only show the x axis labels on the last plot. I also removed the mapping to the names of the days - this is done now via the plot directly. Obviously can be put back in if needed for other reasons!

import pandas as pd
import matplotlib.pyplot as plt
import calendar

df = pd.read_csv("health.csv", header = None, names = ['Physical', 'Emotional'])
# Get Dayofweek index number (start with 6 for sunday) 6,0,1....
df['DayOfTheWeek'] = [(i+6) % 7  for i in range(len(df))]

df_calendar = calendar.Calendar(firstweekday=6)

weeks = int(round(len(df)/7))
fig, axes = plt.subplots(weeks, 1, figsize=(6, weeks*3))

# Loop through the df (splitting week by week)
for i in range(weeks):
    ax=axes[i]

    df.iloc[i*7:(i+1)*7].set_index('DayOfTheWeek').plot(kind='bar', ax=axes[i])
    ax.set_ylim([0,10])
    ax.set_xlim([-0.5,6.5])
    ax.set_xticks(range(7))

    if i == 0:
        ax.legend().set_visible(True)
    else:
        ax.legend().set_visible(False)

    if i == weeks-1:
        ax.set_xticklabels([calendar.day_name[weekday] for weekday in df_calendar.iterweekdays()])
        ax.set_xlabel("Day of the week")
    else:
        ax.set_xticklabels([])
        ax.set_xlabel("")

plt.savefig("health.png")
plt.show()

health

user_22361500
Related