sampleBy (DataFrameStatFunctions)

Returns a stratified sample without replacement based on the fraction given on each stratum.

Syntax

sampleBy(col, fractions, seed=None)

Parameters

Parameter	Type	Description
col	str	The column that defines strata.
fractions	dict	The sampling fraction for each stratum. Strata not specified are treated as having a fraction of zero.
seed	int, optional	Random seed.

Returns

DataFrame

Examples

from pyspark.sql import functions as sf
dataset = spark.range(0, 100, 1, 5).select((sf.col("id") % 3).alias("key"))
sampled = dataset.stat.sampleBy("key", fractions={0: 0.1, 1: 0.2}, seed=0)
sampled.groupBy("key").count().orderBy("key").show()
# +---+-----+
# |key|count|
# +---+-----+
# |  0|    4|
# |  1|    9|
# +---+-----+