232 The ATLAS Collaboration / Physics Letters B 759 (2016) 229–246
Fig. 4. Dijet mass spectra overlaid with the fits to the background function together with the results from BumpHunter and benchmark signals scaled by a factor of 50. The
most discrepant region is indicated by the two blue lines. The lower panels show the significances per bin of the data with respect to the background fit, in terms of the
number of standard deviations, considering only the statistical fluctuations. The distributions are shown for the (a) “1b” and (b) “2b” categories. (For interpretation of the
references to colour in this figure legend, the reader is referred to the web version of this article.)
outcomes is evaluated using the ensemble of Poisson outcomes
across all intervals scanned, by applying the algorithm to many
pseudo-data samples drawn randomly from the background fit.
Without including systematic uncertainties, the probability that
fluctuations of the background model would produce excesses at
least as significant as those observed in the data, anywhere in the
distribution, is greater than 60% in the “1b” and “2b” categories.
Thus, there is no evidence of localized contributions to the mass
distribution from BSM phenomena.
6. Systematic uncertainties
Uncertainties in the parameters of the fitted background func-
tion
Eq. (1) are evaluated by fitting the ansatz to pseudo-data
drawn via Poisson fluctuations around the fitted background
model. The uncertainty in the prediction in each m
jj
bin is taken
to be the root mean square of the function value for 10000 gen-
erated
pseudo-experiments. To estimate an uncertainty due to the
choice of background parameterization, one additional degree of
freedom, z
p
4
log(z)
, is appended as a multiplicative factor to the
nominal ansatz (Eq. (1)), and the difference between the estimated
parameters from the two fits is taken as an uncertainty.
The
uncertainty in the jet energy scale is estimated using vari-
ous
methods in 8TeVdata, corrected to the new centre-of-mass
energy by taking the difference between the 8 TeV and 13 TeV
runs
into account using MC simulation [28]. The jet energy scale
uncertainty used in this analysis relies on a set of three nuisance
parameters [34]. For untagged jets it is within the range 1–5% for
jet transverse momenta greater than 200 GeV.
The
relative additional uncertainty in the energy scale of
b-tagged jets is estimated using the MC samples and verified with
data following the method described in Ref. [35]. The ratio r
trk
of
the sum of track transverse momenta inside the jet to the total
jet transverse momentum measured in the calorimeter is used for
this estimate. The double ratio of r
trk
from data and simulation is
formed and compared for inclusive jets and b-jets. The estimated
relative additional uncertainty for jets with 200 < p
T
< 800 GeV is
found to be less than 2.6%, and this value is subsequently used in
the higher p
T
regions. This relative uncertainty is applied in ad-
dition
to the nominal jet energy scale uncertainty. The maximum
uncertainty for b-tagged jets is estimated to be 6% and is conser-
vatively
applied to all p
T
regions.
The
uncertainty in the jet energy resolution is estimated using
the same method as the untagged jet energy scale uncertainty and
relies on an additional Gaussian smearing of the reconstructed jet
energies in MC simulation. For jets with p
T
> 50 GeV, the uncer-
tainty
is less than 2%.
The
uncertainty introduced by the application of the b-tagging
algorithm is the largest systematic uncertainty in the analysis. The
uncertainty in the measured tagging efficiency of b-jets is esti-
mated
by studying t
¯
t events in 13 TeV data for jet p
T
up to
200 GeV [31]. The uncertainties in the measured rate of mistag-
ging
c-jets and light-flavour jets are estimated in 8TeVdata. The
uncertainties are extrapolated to 13 TeV, taking into account the
addition of the new IBL system as well as reconstruction and tag-
ging
improvements. An additional term is included to extrapolate
the measured uncertainties to the high-p
T
region of interest. This
term is calculated from simulated events by considering varia-
tions
on the quantities affecting the b-tagging performance such
as the impact parameter resolution, percentage of poorly measured
tracks, description of the detector material, and track multiplicity
per jet. The dominant effect on the uncertainty when extrapolat-
ing
at high-p
T
is related to the different tagging efficiency when
smearing the tracks impact parameters based on the resolution
measured in data and simulation. The difference in the impact pa-
rameter
resolution is due to effects from alignment, dead modules
and additional material not properly modelled in the simulation.
The impact of the b-tagging efficiency uncertainty increases with
jet p
T
and reaches 50% above 2TeV.