One of the plausible theories is that dark matter could interact with visible matter and decay into a two-jet signature called a dijet. ATLAS has developed a broad physics program dedicated to various dijet searches, but very few look at the low-mass region. The low-mass region is known to be difficult to probe because the events of interest are produced alongside a large number of normal physics events. The collision rate of the LHC is so high that even the ATLAS Detector cannot keep up with it and thus has to selectively record only a small number of available events of interest. To combat this problem, a new method has been developed whereby only a fraction of the information available in a given collision is recorded. Consequently, we can record more events of interest for the same data bandwidth. These objects can then be used for physics, and the process of doing so is referred to as a Trigger Level Analysis (TLA).

To conduct the analysis and search for new physics, we have to compare this enormous amount of data with a background estimation. This is typically done using a Monte Carlo (MC) simulation. Still, such simulations require huge computing resources, especially given that tens of billions of data points would be needed to have a similar amount of simulated data to compare with the experimental data. This is certainly not feasible and, more importantly, is not sustainable.

The study performed in this thesis focuses on an effort to find a complementary way to perform background estimation in the face of an enormous experimental dataset. First, a data-like shape is created via a fit using functional forms. Then, Poisson-varied sets of pseudodata are randomly generated. This ensemble of pseudodata is then evaluated via different tests, such as goodness-of-fit, spurious signal, signal injection, and background stability tests. A new software framework has been written to perform this task. The result is presented and has been found to be a viable alternative method to generate a robust background expectation, even in the presence of an unprecedented amount of experimental data.