As comments from Mike Lisa in EPD meeting on May. 28,   that directed flow (v1) may flip sign within EPD region for 7.2 GeV data, so it is good to apply \eta weighing and v1 itself can play as the weighting factor 
https://drupal.star.bnl.gov/STAR/blog/lisa/optimizing-ep1-resolution-au27au-ring-dependent-weights-eta-dependent-weights

I applied this method to see how much it can improve the EPD 1st order event planes

Outline:

1. Method
2. Event planes that I use
3. Result
4. Summary
5. Next step

1. Method

2. Event planes that I use

I use 4 EPD-sub event planes (eta range: {-5.16,-3.82,-3.28,-2.87,-2.60} to set 4 EPD-sub region ) and one TPC event plane (all the eta region, will choose a smaller eta region to reduce the effect of momentum conservation). A plot of the EPD-sub coverage,

Figure 1: EPD-sub event planes tile coverages

To get an idea of the \eta range for each supersector, plot TnMIP vs. supersepctor VS \eta,

Figure 2: \eta range of each and every supersectors for FXT 7.2 GeV

I use EPD-3 as my primary event plane to calculate v1 and only for hits in EPD-3 region, I use EPD-4 instead to reduce the effect of auto-correlation

3. Results

After several iteration, the shape of v1 VS eta become stable. Plots from 9 centrality bins (0-5%, 5-10%, 10-20%, 20-30%, 30-40%, 40-50%, 50-60%, 60-70%, 70-80%),

Figure 3: v1 <\phi - \psi_{1}^{EPD-3, or EPD-4}> VS \eta in each and every centrality bins. Note that the POL3 fitting should be corrected with \eta_CM rather than y_CM that I am using here by mistake.
4. Summary

• v1 are small (close to zero) in forward region.
• In mid & peripheral collisions, v1 in EPD region cross 0.
• Around \eta = -2 (gap of TPC, EPD), v1 are not smooth
• POL3 doesn't fit well.

5. Next step

• Check the correctness of my analysis: How I fill histograms, the calculation of EPs, other bugs... (to be continue ...)

References:
https://drupal.star.bnl.gov/STAR/blog/dchen/epd-event-planes-snn-3-72-gev-fxt-collisions