Value Search FOR deviations/outliers?

[Ben]

I'm trying to create a Value Search to only return deviations/outliers in my dataset, but can't think of a way to do so. For instance, the Value Search for the image below would return the date ranges of Jan 1 - Jan 11 and Apr 6 - May 11.

I've tried creating a Reference Profile and Custom Conditions, but it's tricky because this variable can move up or down "smoothly", while these violent deviations/outliers can show up at any point.

Most documentation and forum threads seem to cover filtering outliers, and not identifying them. Any help would be appreciated. Thanks!

Edited May 17, 2021 by Ben

[Teddy]

Ben,

You can probably use a moving standard deviation to find that condition.

Below is an example formula. Note: you can use signal from condition too, but you may want to create a condition with the periods function so you can have better control over the moving range:

$signal.aggregate(stdDev(), periods(20min,5min), middleKey())

Hope this helps.

Teddy

[Thorsten Vogt]

Hi Ben,

another option is using the function highpassfilter() as in the following example:

$signal.highPassFilter(20min, 2min, 60).abs().agileFilter(2min)

The resulting signal can then be used in Value Search to determine the deviations.

Regards,

Thorsten