python - Multi-line chart with seaborn tsplot

Question

I want to create a smoothed line chart using matplotlib and seaborn.

This is my dataframe df:

hour    direction    hourly_avg_count
0       1            20
1       1            22
2       1            21
3       1            21
..      ...          ...
24      1            15
0       2            24
1       2            28
...     ...          ...

The line chart should contain two lines, one for direction equal to 1, another for direction equal to 2. The X axis is hour and Y axis is hourly_avg_count.

I tried this, but I cannot see the lines.

import pandas as pd
import seaborn as sns
import matplotlib
import matplotlib.pyplot as plt

plt.figure(figsize=(12,8))
sns.tsplot(df, time='hour', condition='direction', value='hourly_avg_count')

Answer 1

The tsplot is a bit strange or at least strangly documented. If a dataframe is supplied to it, it assumes that there must be a unit and a time column present, as it internally pivots about those two. To use tsplot to plot several time series you would therefore need to supply an argument to unit as well; this can be the same as condition.

sns.tsplot(df, time='hour', unit = "direction", 
               condition='direction', value='hourly_avg_count')

Complete example:

import numpy as np
import pandas as pd
import seaborn as sns
import matplotlib.pyplot as plt

hour, direction = np.meshgrid(np.arange(24), np.arange(1,3))
df = pd.DataFrame({"hour": hour.flatten(), "direction": direction.flatten()})
df["hourly_avg_count"] = np.random.randint(14,30, size=len(df))

plt.figure(figsize=(12,8))
sns.tsplot(df, time='hour', unit = "direction", 
               condition='direction', value='hourly_avg_count')

plt.show()

Also worth noting that tsplot is deprecated as of seaborn version 0.8. It might thus be worth to use some other way to plot the data anyways.