Standard deviation of TrackedVariable

[jeremy_r]

TrackedVariables have a built-in "average" command that calculates the average for STAT_TYPE_LEVEL (weighted by time) and STAT_TYPE_TIME_SERIES (unweighted) TrackedVariables.

It would be really helpful if these two types of TrackedVariable also supported calculating the Standard Deviation of the data points.

[joerg_vogel_HsH]

You don’t know which statistical distribution is the base for the tracked events. A standard or normal distribution needs some conditions that you can compute the deviation. You can’t use this distribution, if the intervals of the events or the level values are restricted to few.

[jeremy_r]

For unweighted values, you should just be able to choose between the sample standard deviation and the population standard deviation, like in SQL.

For weighted values I'm less sure of, but it appears to me that for STAT_TYPE_LEVEL the durations would represent frequency weights, so it would be appropriate to use a frequency-weighted distribution (https://en.wikipedia.org/wiki/Weighted_arithmetic_mean#Frequency_weights).