1
Vote

.ps files are not rotated properly

description

For exampmle - these guys all came in virtical, rather than horizontal.
Hi Volker, Marco,
 
Because we don't have enough time / knowledge to do MC study on it in short time scale, so here is a proposal to evaluate this systematics from data.
 
What I did is comparing pre-tag and tagged event and check if unknown shape difference we could see.
Here is a plot, ratio of pre-tag to single tag on dijet mass:
http://www-clued0.fnal.gov/~enari/d0_private/dijet_sys/ratio_pre_st.ps
Black filled points are data, open points are MC (including QCD).
 
Then take a ratio between two, the plot is http://www-clued0.fnal.gov/~enari/d0_private/dijet_sys/double_ratio_pre_st.ps
<double_ratio_pre_st.ps>
Dot is the ratio, and blue and red lines are uncertainties from ALPGEN (reweighting, kt, MMLM).
 
In this ratio, potentially mis-modeling in bID/JES/JSSR uncertainties could be included which we already assigned uncertainties.
It could be over estimating uncertainties, but only thing we can do now is propagating this shape to final output. I fit the ratio with 3rd pronominal function, I got http://www-clued0.fnal.gov/~enari/d0_private/dijet_sys/double_ratio_pre_st_fit.ps
<double_ratio_pre_st_fit.ps>
 
Then I apply this function as weight and check effect on NN output:
http://www-clued0.fnal.gov/~enari/d0_private/dijet_sys/
<double_ratio_pre_st.ps>NNoutput_after_dijet_rew.ps
 
1st page is for single tag, and 2nd page is for double tag. At the end, this is not so big shape uncertainty.
 
Please let me know if this procedure works and looks reasonable to evaluate systematic uncertainty on heavy flavor MC modeling. If you agree, I'll put them also in the note.
 
By the way, I can quickly make collie input, but I'm not sure we can update it again.
This will be another discussion within the conveners...
 
Regards,
Yuji for WH team.

comments