Get top rows or random rows within groups in a DataFrame

# Import libraries
import pandas as pd
import seaborn as sns

# Load sample data in a DataFrame
df = (
    sns.load_dataset('iris')
    .sample(n=20, random_state=20)
    .sort_values('species')
    .reset_index(drop=True)
    [['species', 'sepal_width']]
)
df

	species	sepal_width
0	setosa	3.2
1	setosa	3.0
2	setosa	3.4
3	setosa	3.2
4	setosa	3.7
5	setosa	3.0
6	versicolor	2.7
7	versicolor	2.8
8	versicolor	2.9
9	versicolor	2.8
10	versicolor	2.9
11	versicolor	2.5
12	versicolor	2.4
13	virginica	3.2
14	virginica	2.5
15	virginica	2.8
16	virginica	3.0
17	virginica	2.2
18	virginica	3.0
19	virginica	2.8

Get top N rows of each group

An option is to sort values, then use groupby() followed by head()

# Get top 3 rows for each group, sorted by decreasing sepal width
(
    df
    .sort_values(['species', 'sepal_width'], ascending=[True, False])
    .groupby('species')
    .head(3)
)

	species	sepal_width
4	setosa	3.7
2	setosa	3.4
0	setosa	3.2
8	versicolor	2.9
10	versicolor	2.9
7	versicolor	2.8
13	virginica	3.2
16	virginica	3.0
18	virginica	3.0

Get random rows within each group

To retrieve n random rows from each group, use sample(n) after a groupby():

# Get 1 random row within each group of the DataFrame
(
    df
    .groupby('species')
    .apply(lambda x: x.sample(1))
)

		species	sepal_width
species
setosa	1	setosa	3.0
versicolor	10	versicolor	2.9
virginica	13	virginica	3.2