Plot Dataframe doesn't start from the beginning with matplotlib.axes.plot()

user_14221439 0
Viewed 266

I got a dataframe Filled_series:

2015-05-03     0.00000
2015-05-09         NaN
2015-05-15     7.27943
2015-05-21         NaN
2015-05-27     7.26766
                ...   
2019-10-03    12.96608
2019-10-09     9.42112
2019-10-15     6.36359
2019-10-21    10.27396
2019-10-27     9.76496
Name: 367, Length: 274, dtype: float32

and I want to plot it so I do :

fig, ax = plt.subplots(figsize=(15,15))
x_ax = Filled_series.index
ax.plot(x_ax, Filled_series)

Here is what I got enter image description here

With ax.scatter it works fine but not with ax.plot

It doesn't start from the 2015-05-03, I don't know why. How can I do to get the whole time series ?

pandas matplotlib python
1 Answers

The plotting works fine, but the plotting function can't plot a line between a point and a NaN. If you force matplotlib to plot markers (e.g. by line style -o), you can see that it plots the non-NaN-points but won't produce a line from them to a NaN. Lines are only drawn between existing points.

This is why you have this wide withe space at the beginning of your plot. The points are there (and therefore, the axis's limits respect them) but you can't see them because per default, the line-plot plots lines and not points (aka. markers). This is also the reason why the scatter-plot works. It draws points rather than lines.

Perhaps, it gets a bit clearer, when looking at an example:

import pandas as pd
import matplotlib.pyplot as plt
import datetime
import numpy as np

data = [['2015-05-03',     0.00000],
['2015-05-09',     np.nan ],
['2015-05-15',     7.27943],
['2015-05-21',     np.nan ],
['2015-05-27',     7.26766],
['2015-06-03',     np.pi]]

df = pd.DataFrame(data,columns=['date_str','val'])
# convert to datetime
df['date'] = [datetime.datetime.strptime(s, '%Y-%m-%d') for s in df['date_str']]

df.plot(kind='line',x='date',y='val',rot=45,style='-o')

output fig

So now, if you want to ignore the NaN-entries, you can mask them:

mask = ~np.isnan(df['val'])
df[mask].plot(kind='line',x='date',y='val',rot=45)

ignore NaNs

BTW, the pandas.DataFrame.plot() method uses matplotlib.pyplot.plot() in the background. It is sometimes a bit more convenient because it adds labels and legends directly^^

user_124701960
Related