You signed in with another tab or window. Reload to refresh your session.You signed out in another tab or window. Reload to refresh your session.You switched accounts on another tab or window. Reload to refresh your session.Dismiss alert
For a cost-benefit analysis I need to know the value which covers in 99%/95% of cases. In my case the highest value which covers those cases after grouping.
Potential Solutions:
Looking at the median implementation, It seems like it would be quite easy to implement given that the values are pre-sorted, and I'm willing to implement it and create a PR.
But I'm wondering what the best way would be to implement it since it has the second-parameter percentile. Adding another 'weighted-sum' type seems like it would be quite a lot of additional items on the UI side.
Additionally you would probably want a max and a min variant depending on the underlying values & analysis.
I will probably implement a proof of concept for my own project, but if there is a preferred way to do that would help.
The text was updated successfully, but these errors were encountered:
Feature Request
Add aggregator for percetiles (e.g. 99pt, 95pt).
Description of Problem:
For a cost-benefit analysis I need to know the value which covers in 99%/95% of cases. In my case the highest value which covers those cases after grouping.
Potential Solutions:
Looking at the median implementation, It seems like it would be quite easy to implement given that the values are pre-sorted, and I'm willing to implement it and create a PR.
But I'm wondering what the best way would be to implement it since it has the second-parameter percentile. Adding another 'weighted-sum' type seems like it would be quite a lot of additional items on the UI side.
Additionally you would probably want a max and a min variant depending on the underlying values & analysis.
I will probably implement a proof of concept for my own project, but if there is a preferred way to do that would help.
The text was updated successfully, but these errors were encountered: