Github user brkyvz commented on a diff in the pull request:
https://github.com/apache/spark/pull/5858#discussion_r29557546
--- Diff:
sql/core/src/main/scala/org/apache/spark/sql/execution/stat/StatFunctions.scala
---
@@ -23,29 +23,43 @@ import org.apache.spark.sql.types.{DoubleType,
NumericType}
private[sql] object StatFunctions {
+ /** Calculate the Pearson Correlation Coefficient for the given columns
*/
+ private[sql] def pearsonCorrelation(df: DataFrame, cols: Seq[String]):
Double = {
+ val counts = collectStatisticalData(df, cols)
+ counts.Ck / math.sqrt(counts.MkX * counts.MkY)
+ }
+
/** Helper class to simplify tracking and merging counts. */
private class CovarianceCounter extends Serializable {
- var xAvg = 0.0
- var yAvg = 0.0
- var Ck = 0.0
- var count = 0L
+ var xAvg = 0.0 // the mean of all examples seen so far in col1
+ var yAvg = 0.0 // the mean of all examples seen so far in col2
+ var Ck = 0.0 // the co-moment after k examples
+ var MkX = 0.0 // sum of squares of differences from the (current) mean
for col1
+ var MkY = 0.0 // sum of squares of differences from the (current) mean
for col1
+ var count = 0L // count of observed examples
// add an example to the calculation
def add(x: Double, y: Double): this.type = {
- val oldX = xAvg
+ val deltaX = x - xAvg
+ val deltaY = y - yAvg
count += 1
- xAvg += (x - xAvg) / count
- yAvg += (y - yAvg) / count
- Ck += (y - yAvg) * (x - oldX)
+ xAvg += deltaX / count
+ yAvg += deltaY / count
+ Ck += deltaX * (y - yAvg)
+ MkX += deltaX * (x - xAvg)
+ MkY += deltaY * (y - yAvg)
--- End diff --
Same here
---
If your project is set up for it, you can reply to this email and have your
reply appear on GitHub as well. If your project does not have this feature
enabled and wishes so, or if the feature is enabled but not working, please
contact infrastructure at infrastruct...@apache.org or file a JIRA ticket
with INFRA.
---
---------------------------------------------------------------------
To unsubscribe, e-mail: reviews-unsubscr...@spark.apache.org
For additional commands, e-mail: reviews-h...@spark.apache.org
