backtrack / masking

Hey i try to outline different backtrack and masking ideas and the problems we could run into. In each example i just show the backtrack of one single variable/data-series, of course we will have multiple of them, and may or may not store the backtrack at a different location, but these are implementation details, we don't want to bother us here.

⛔ To keep this discussion clean, because its such delicate and complicated matter, i will post each proposal in a own comment. Please use the reply to comment function , if the comment is to a special proposal..

This will be a living document, so i will add more proposals, but it may need some time, to get it all right.. cheers b

Nomenclature

• BAD = 99
• DOUBT = 50
• UNFLAGGED = 0
• flagged: to a data point, a flag was attached, by the rules that the user(?) or we (?) define. eg. new > old
• unflagged : the test has currently the flag 0
• touched: a test like to flag a certain point, nevertheless it already flagged, and nevertheless the flag there (if any) is worse/better/equal...
• untouched: the data point was never touched by any test (or at least look like this)
• overwrite policy: how flags gets overwritten by others. We somehow want the user to define the policy (eg. new > old or new >= old or whatever..
• BT: backtrack, the blob where we dump our precious stuff, to fiddle and shrobble with it later, if we have enough kittns togetha..