search

umd-lhcb / hammer-reweight

Code for FF reweighting in HAMMER
BSD 2-Clause "Simplified" License
0 stars 0 forks source link

Validation of HAMMER reweighting with known FF distributions with given parameterizations #3

Closed yipengsun closed 3 years ago

yipengsun commented 3 years ago

Here I'll record more details about validation mentioned in umd-lhcb/lhcb-ntuples-gen#51.

yipengsun commented 3 years ago

In Manuel's BaBar code (plot_q2_2hdm.cxx), he did not use TF1; instead, he filled a TH1F histogram with explicit binning and q2 calculated at center of the bin.

yipengsun commented 3 years ago

Here's some non-working plot. The main problem now is the normalization: ff_cln_reweighted

I also tried to compute total number of entries with histogram.GetIntegral(), and the result is 0. This is likely due to each bin content is too small (1E-15)

yipengsun commented 3 years ago

@manuelfs I've updated a plot that's been heavily tweaked in umd-lhcb/lhcb-ntuples-gen#51. Here I want to offer a little explanation on why this takes ~1 week to finish.

  1. Doesn't start until Wednesday
  2. I used a naive histo.Scale(1/histo.Integral()), and I got something like this: scaled_by_1
  3. I didn't understand why I seemed to be getting negative components, so I started reading documentation
  4. During the next 2 days, I did a few experiments with TH1D, and it seems that my method was mostly right (sans the option width)
  5. I was confused for a bit, until today, where I really sat down and looked closely to the documentation
  6. I realized that the visually negative components are just error bars.
yipengsun commented 3 years ago

We consider this validated. The thing to keep in mind is that it might be useful to make reference curves binned in the same way as data/MC to make comparison easier.

To do that, one needs to integrate the continuous function inside each bin.