Stats

Streaming statistics for time series data.

The stdlib stats module offers "standardization" of metric time series -- shifting and re-scaling based on estimates of their center and scale. This places streams into a standard form for comparison with others, or for thresholding based on deviations from the center. Besides "z-score" standardization using the sample mean and estimated standard deviation, there is a robust variant (MAD) and mean-relative (CV).

All stats methods are reducers, and therefore can only be used in stream context.

demean - reducer

Return the difference between the last observed value and the batch average.

... | reduce value=demean(value) | ...
Option Description Required?
-field name of the field to apply the reduce operation on Yes

z - reducer

Return the sample Z-score of the specified field.

... | reduce value=z(value) | ...
Option Description Required?
-field name of the field to apply the reduce operation on Yes

relMean - reducer

Return the last observed value as a percentage of the batch mean

... | reduce value=relMean(value) | ...
Option Description Required?
-field name of the field to apply the reduce operation on Yes

cv - reducer

Return the coefficient of variation (stdev / mean) of the batch

... | reduce value=cv(value) | ...
Option Description Required?
-field name of the field to apply the reduce operation on Yes