How to change the point size for regplot(), seaborn's scatter plot function (python)

Question

I want to be able to set the point size when plotting like this:

sns.regplot(y=[1,3,4,2,5], x=[range(5)], data=df,
            marker='o', color='red')
plt.show()

Do you guys know how?

Answer 1

To do this you can feed the regplot() function the scatter_kws arg like so:

import seaborn as sns
tips = sns.load_dataset('tips')
sns.regplot(x='total_bill', y='tip', data=tips,
            marker='o', color='red', scatter_kws={'s':2})

sns.regplot(x='total_bill', y='tip', data=tips,
            marker='o', color='red', scatter_kws={'s':20})

Answer 2

You could even make the points dynamically sized to represent a third dimension. This code uses the same data as the OP, but wraps it in a DataFrame (as seaborn is designed for that) and also adds a third dimension, z.

import seaborn as sns
import pandas as pd

data = pd.DataFrame({
    'x': [x for x in range(5)],
    'y': [1, 3, 4, 2, 5],
    'z': [14, 14, 100, 16, 36]
})
sns.regplot(x='x', y='y', data=data, marker='o', color='red',
    scatter_kws={'s': data['z']})

You can probably imagine how you could also manipulate the list/array of sizes programatically, giving you a lot of power to convey extra information.

If you wanted to split these up into two groups, like in a FacetGrid, regplot doesn't handle that keyword mapping cleanly. But scatterplot does. Just plot regplot underneath for the fit line, then scatterplot over it for the points. Below, I've added a grp variable and plotted it onto a FacetGrid.

import seaborn as sns
import pandas as pd

data = pd.DataFrame({
    'x': [x for x in range(5)],
    'y': [1, 3, 4, 2, 5],
    'z': [14, 14, 100, 16, 36],
    'grp': ['a', 'a', 'b', 'b', 'a'],
})
g = sns.FacetGrid(data=data, col='grp', margin_titles=True)
g.map_dataframe(sns.regplot, 'x', 'y', color='red', scatter=False)
g.map_dataframe(sns.scatterplot, 'x', 'y', size='z', marker='o', color='red')

Answer 3

I would add to mburke05's answer that it appears possible to pass array-like data into scatter_kws. For example, if you wanted the size attribute in the tips dataset to determine the size of a point you can write:

sns.regplot(
    x="total_bill", y="tip", data=tips,
    marker='o', color='red', scatter_kws={'s':tips['size']})

However, you must explicitly lookup that attribute in the dataframe (as above); you cannot simply use the column name as you would when setting x and y.