What inspired this wish list request?
I use column groupings a lot due to wide data sets but when I have down selected factors I would like to use in later analysis I need to be able to identify the factors easily. Currently this requires breaking the original grouping structure to create a new grouping that has only the factors of interest. This makes the data table less organized and harder to use for other analysis.
What is the improvement you would like to see?
I would like the ability to have virtual groupings where a custom grouping can be made but will leave the original column groups and column order as they are. So in a use case, I could have my X's in a virtual column that I could drag into an analysis platform but my original groupings and column order is not affect. The virtual columns would not show up in the data, only in column list boxes.
Icing on the cake would be for any multi-variable modeling platform to allow selected factors in either access factor importance or column contribution charts to be directly added to a virtual grouping.
Why is this idea important?
This will help make analysis faster while still keeping the organization of a data table intact.