Open SteveOv opened 5 months ago
testing revised criteria PDC_TOT - (PDC_NOI if PDC_NOI> 0.6 else 1) (descending) on an overnight run.
effectively the difference between PDC_TOT & PDC_NOI with an extra penalty for PDC_NOI that's low or missing. If it's missing we can't trust it as we don't know what it is. I've seen some LCs with highly distorted LC and these have correlated with low (<0.6) PDC_NOI.
Having problems with TIC197887879 and potentially 2 or 3 others. There are 7 sectors and all bar one (sector 4) have PDC_NOI > 0.99. The current algorithm will equally penalize the remaining 6 sectors. We end up using sectors 2, 3, 4, 31 and 69 however sectors 2 and 69 are the least likely to have a usable signal.
We need to modify the use of the PDC_NOI metric to further penalize these sectors; perhaps the penalty should take the value of PDC_NOI into account.