- Methods:
   a. Simple division only (= normal corr/event mixed corr) -> suffix "div"
   b. Division after self normalize each of them (= normal corr after self norm / event mixed corr afer self norm) -> suffix "norm"

- Self nomalization: for a correlation TH2 (FPS slat vs. FMS points position projected),
   1. Get X projection: then each bin corresponds to sum of all FPS slats along the position.
   2. Normalize 'entire slats contributed to the sum' by sum itself.
   -> By doing this signal (stands out correlation) also will be a little scaled down,
        but fluctuating backgrounds in other slats will be effectively suppressed.
   -> Also peak position itself cannot be changed (?)
   -> Then two factors are dealt with simultaneously
        1) Concentration of FMS points in certain position (FMS cell's efficiency)
        2) FPS slats' efficiency (= effectiveness/accuracy of FPS Mip)

- I expect method a. and b. will give similar peak position if both FMS and FPS are well calibrated
   -> Therefore checked a few RUN15 runs: it seems both results are not very different compared to RUN17 runs
   * this could be intrinsic difference between RUN15 (200 GeV) and RUN17 (510 GeV)

- Concern about division method: I cannot see FMS-FPost hadron correlation not at all by division