Strange calculation for Variance and StandardDeviation

[larshagencognite]

Type of issue

• Bug
• Enhancement
• Compliance
• Question
• Help wanted

Current Behavior

The calculation of standard deviation and variance in https://github.com/OPCFoundation/UA-.NETStandard/blob/master/Libraries/Opc.Ua.Server/Aggregates/StdDevAggregateCalculator.cs is very strange.

• Bounds are always included for variance, never for standard deviation
• The square root is taken on the result for VariancePopulation, StandardDeviationPopulation, and not for the sample variants.
• When including bounds, the sum of squared errors is divided by n+1, I would expect n-1.
• When including bounds, the last bound is ignored.
• When not including bounds, the last data point in the range is ignored.

Expected Behavior

I would expect the calculations to follow the documentation.

• StandardDeviation* = sqrt(Variance*)
• Don't ignore last value
• Divide by n-1 for samples

Steps To Reproduce

No response

Environment

No response

Anything else?

No response

[ThomasNehring]

the fact that the standard deviation is not always the sqrt of the Variance is an error and probably due to a typo when calling the method ComputeValue(TimeSlice).

I assume that dividing by n+1 instead of dividing by n-1 is a typo, also. (see, e.g. Why we use (n.1) in the denominator of standard deviation formula.)

For the question of when to include the bounds I need to have a closer look at the pertinent sections of the spec.
StandardDeviationSample
VarianceSample
StandardDeviationPopulation
VariancePopulation

Updated (202402071607 CET):
Actually, there is sample data in the spec (part 13, section A.1), together with the expected outcome of the aggregate calculation (sections A.35, A.36, A.37, A.38) which we can use to figure out whether the behavior wrt to the boundary values is in sync with the specification.