I have a pandas dataframe and would like to plot values from one column versus the values from another column. Fortunately, there is plot method associated with the dataframes that seems to do what I need:
df.plot(x='col_name_1', y='col_name_2')
Unfortunately, it looks like among the plot styles (listed here after the kind parameter), there are not points. I can use lines or bars or even density but not points. Is there a work around that can help to solve this problem?
You can specify the style of the plotted line when calling df.plot:
df.plot(x='col_name_1', y='col_name_2', style='o')
The style argument can also be a dict or list, e.g.:
import numpy as np
import pandas as pd
d = {'one' : np.random.rand(10),
'two' : np.random.rand(10)}
df = pd.DataFrame(d)
df.plot(style=['o','rx'])
All the accepted style formats are listed in the documentation of matplotlib.pyplot.plot.
dtypes of the dataframe and convert it from object to numeric columns. This issue wasted at least a few hours for me.For this (and most plotting) I would not rely on the Pandas wrappers to matplotlib. Instead, just use matplotlib directly:
import matplotlib.pyplot as plt
plt.scatter(df['col_name_1'], df['col_name_2'])
plt.show() # Depending on whether you use IPython or interactive mode, etc.
and remember that you can access a NumPy array of the column's values with df.col_name_1.values for example.
I ran into trouble using this with Pandas default plotting in the case of a column of Timestamp values with millisecond precision. In trying to convert the objects to datetime64 type, I also discovered a nasty issue: < Pandas gives incorrect result when asking if Timestamp column values have attr astype >.
Pandas uses matplotlib as a library for basic plots. The easiest way in your case will using the following:
import pandas as pd
import numpy as np
#creating sample data
sample_data={'col_name_1':np.random.rand(20),
'col_name_2': np.random.rand(20)}
df= pd.DataFrame(sample_data)
df.plot(x='col_name_1', y='col_name_2', style='o')
However, I would recommend to use seaborn as an alternative solution if you want have more customized plots while not going into the basic level of matplotlib. In this case you the solution will be following:
import pandas as pd
import seaborn as sns
import numpy as np
#creating sample data
sample_data={'col_name_1':np.random.rand(20),
'col_name_2': np.random.rand(20)}
df= pd.DataFrame(sample_data)
sns.scatterplot(x="col_name_1", y="col_name_2", data=df)
Now in latest pandas you can directly use df.plot.scatter function
df = pd.DataFrame([[5.1, 3.5, 0], [4.9, 3.0, 0], [7.0, 3.2, 1],
[6.4, 3.2, 1], [5.9, 3.0, 2]],
columns=['length', 'width', 'species'])
ax1 = df.plot.scatter(x='length',
y='width',
c='DarkBlue')
https://pandas.pydata.org/pandas-docs/version/0.23/generated/pandas.DataFrame.plot.scatter.html
When this question was posted, scatter plot was a separate function in pandas. Since pandas 0.13, you can use kind='scatter' to plot a scatter plot from two columns.
df = pd.DataFrame({'colA': np.random.rand(10), 'colB': np.random.rand(10)})
df.plot(x='colA', y='colB', kind='scatter')
If you want to change the marker (e.g. x), then you can use marker= parameter:
df.plot(x='colA', y='colB', kind='scatter', marker='x')
df.plot(style='o')?Under the hood, df.plot defaults to a matplotlib line plot, (i.e. Axes.plot() or plt.plot), so passing style= is similar to plt.plot(x, y, 'o'). In particular, this creates an Axes.lines object which stores marker attributes.
On the other hand, df.plot(kind='scatter') (or df.plot.scatter) uses Axes.scatter of matplotlib; this creates an Axes.collections object to store marker attributes.
One significant difference is when you want to change marker size; with kind='scatter', you have to use s= like plt.scatter but with a line plot, you have to use ms= instead. The following two function calls produce the same output.
df.plot(x='colA', y='colB', kind='scatter', s=36) # case 1
df.plot(x='colA', y='colB', style='o', ms=6) # case 2
You can read this Q/A about why marker size values should be different in order for these methods to produce the same output.