Absorbing systematic effects to obtain a better background model in a search for new physics

This paper presents a novel approach to estimate the Standard Model backgrounds based on modifying Monte Carlo predictions within their systematic uncertainties. The improved background model is obtained by altering the original predictions with successively more complex correction functions in signal-free control selections. Statistical tests indicate when sufficient compatibility with data is reached. In this way, systematic effects are absorbed into the new background model. The same correction is then applied on the Monte Carlo prediction in the signal region. Comparing this method to other background estimation techniques shows improvements with respect to statistical and systematical uncertainties. The proposed method can also be applied in other fields beyond high energy physics.

💡 Research Summary

The paper introduces a unified framework for improving Standard Model background estimates in searches for new physics by “absorbing” systematic effects directly into the Monte Carlo (MC) prediction. Traditional background modeling relies on MC simulations with separate, often conservative, systematic uncertainties for each process. Because these uncertainties are typically correlated, non‑linear, and difficult to propagate, the resulting background can be mismatched to data, limiting the sensitivity of searches.

The authors’ method proceeds in four logical stages. First, a signal‑free control region (CR) is defined – for example a side‑band in invariant mass or a region defined by orthogonal trigger selections – where the presence of new physics is highly unlikely. In this region the observed data are compared directly to the unmodified MC prediction, providing a baseline measure of the MC’s fidelity to the real detector and physics environment.

Second, the MC prediction in the CR is modified by a series of increasingly flexible correction functions. The authors start with the simplest possible model – a global linear scale factor – and, if the data‑MC agreement remains insufficient, they introduce higher‑order polynomial terms, spline interpolations, or locally weighted functions. At each step the parameters of the correction function are obtained by a fit (least‑squares or maximum‑likelihood) that explicitly incorporates the statistical uncertainties of the data.

Third, statistical compatibility is assessed using a battery of goodness‑of‑fit tests (χ², Kolmogorov–Smirnov, Anderson–Darling). An a‑priori significance threshold is set; when the test statistic falls below this threshold the correction is deemed adequate and the algorithm stops, preventing over‑fitting. The covariance matrix of the fitted parameters is retained for later error propagation.

Fourth, the final correction function – the one that achieved sufficient agreement in the CR – is applied unchanged to the MC prediction in the signal region (SR). The uncertainties on the correction parameters are propagated to the SR background estimate, thereby embedding the systematic discrepancy observed in the CR into the SR model. This yields a background that is effectively “data‑driven” while still retaining the full kinematic detail of the MC simulation.

The authors benchmark their approach against conventional techniques such as side‑band scaling, ABCD methods, and pure data‑driven estimations. They find that statistical uncertainties are comparable or modestly reduced, while systematic uncertainties shrink dramatically because the correction function has already absorbed the dominant mismodeling effects. Moreover, the method naturally handles correlations across multiple analysis channels, improving the overall sensitivity of combined searches.

From a computational standpoint, the stepwise increase in function complexity is efficient: the algorithm terminates as soon as the data‑MC agreement is acceptable, avoiding unnecessary high‑dimensional parameter scans. The authors also discuss the method’s robustness against limited statistics in the CR, showing that regularisation (e.g., penalising overly complex functions) can be incorporated without biasing the final result.

Beyond high‑energy physics, the paper argues that any field that relies on sophisticated simulations with sizable systematic uncertainties can benefit from this strategy. Examples include astrophysical background modeling (e.g., diffuse gamma‑ray emission), medical imaging (e.g., tissue‑specific Monte Carlo dose calculations), and environmental modeling (e.g., atmospheric pollutant dispersion). The key requirement is the existence of a control region where the signal is negligible and a well‑defined mapping between simulation and observation can be established.

In conclusion, the study presents a pragmatic, statistically rigorous technique for background estimation that merges the strengths of simulation‑based modeling with data‑driven correction. By absorbing systematic discrepancies in a controlled, test‑driven manner, the method reduces overall uncertainties and enhances the discovery potential of new‑physics searches, while offering a versatile tool applicable across a broad spectrum of scientific disciplines.