Open
Description
The work in #516 and #519 implements new methods for computing impacts but also introduce a more comprehensive stored container with the impacts grouped by modifier types as well as grouped systematic, statistical and normalisation/shape factor impacts.
In these PRs, this container is not exposed to the user and is not tested. This is to keep the PRs modular and prevent them from bloating. In a follow-up PR, the following is needed:
- Return these grouped impacts in the RankingResults object (or an appropriate dedicated object)
- Write tests for grouped impact summing:
- For covariance-based
- For
auxdata
shifting - For NP shifting
- return the impacts via
tabulate
in nice formats - Produce one plot with 2 panels or 2 separate plots or allow the user to choose one out of the 2 but default to 2 of:
- The impacts per modifier, horizontal bar chart style
- The grouped systematic/stat/NFs/SFs impacts (summary impacts)