Given a numeric column within a Spark DataFrame, compute approximate quantiles.

sdf_quantile(
  x,
  column,
  probabilities = c(0, 0.25, 0.5, 0.75, 1),
  relative.error = 1e-05,
  weight.column = NULL
)

Arguments

x

A spark_connection, ml_pipeline, or a tbl_spark.

column

The column(s) for which quantiles should be computed. Multiple columns are only supported in Spark 2.0+.

probabilities

A numeric vector of probabilities, for which quantiles should be computed.

relative.error

The maximal possible difference between the actual percentile of a result and its expected percentile (e.g., if `relative.error` is 0.01 and `probabilities` is 0.95, then any value between the 94th and 96th percentile will be considered an acceptable approximation).

weight.column

If not NULL, then a generalized version of the Greenwald- Khanna algorithm will be run to compute weighted percentiles, with each sample from `column` having a relative weight specified by the corresponding value in `weight.column`. The weights can be considered as relative frequencies of sample data points.