Splunk Search

Can eval be used to calculate the standard deviation in multiple fields for a single event?

Path Finder

Is there a way to use eval to calculate the standard deviation of data in multiple fields (same number of fields each time) for a single event? Right now I am using an eval statement that writes out the entire Standard Deviation formula.

| eval StDev = round(pow((pow(field1-fieldsAvg, 2)+pow(field2-fieldsAvg, 2)+pow(field3-fieldsAvg, 2)+pow(field4-fieldsAvg, 2)+pow(field5-fieldsAvg, 2))/5, 1/2), 4)

Each line of my table represents a different event that needs this calculation.


Since Standard deviation is calculated using average so I am assuming your field called fieldsAvg is the average of all the five fields. Which also makes me feel we can tweak your situation as follows:
- Make a new field called myField which has values from all the five fields. So if you have 3 events with 5 field values each, this new field will have 15 values to take care of all 5 fields for all 3 events.
- Calculate the stdev on this new field

your base query to return field1,field2,field3,field4,field5
| eval myField=mvzip(field1, mvzip(field2, mvzip(field3, mvzip(field4, field5)))) 
| mvexpand myField 
| rex max_match=0 field=myField "(?<numbers>\d+)"
| stats stdev(numbers) as stdDeviation

Path Finder

Yes, I do have the fieldsAvg calculated separately.

In your explanation, you say that 5 values from 3 events will all end up in myField together. What if I still want to keep the Standard Deviations separate by events?

0 Karma


You could improve your current solution by making a macro out of it, which would be easier to use and maintain across different searches without worrying about a typo causing one of them to behave differently.

0 Karma


My understanding is there is no eval function that will calculate the standard deviation for fields in the same row.

Most of the eval functions are designed to be performed across all of the rows against specific fields (e.g., the standard deviation for your field "field1").

0 Karma