Two central questions facing numerous fields will be addressed by this award: "How does one reach rigorous quantitative conclusions when the science involves complicated numerical models and large scale data sets?" and "How can one use computational methods to point out new insights into what are the most important features, parameters and data?". To this end, a generalized statistical and visualization structure will be developed for the purpose of comparing and analyzing large-scale computational models to large heterogenous data sets. The codes, techniques, algorithms and methodologies applied to this project will be applied to critical problems in four different fields: high-energy nuclear physics, astrophysics, meteorology and biology. By devising a flexible open-source framework, the products of this award will be extendable to numerous other problems in the natural or social sciences involving sophisticated simulations and complicated data structures.
The way in which simulations and data are analyzed will be revolutionized by the capabilities developed by this work. Fundamental model parameters related to each of the four problems, such as the percentage of the dark matter in the universe or the viscosity of the quark-gluon plasma, will be stated with rigorous uncertainties. Methods will be developed and implemented for quantitatively stating which data were most important for constraining parameters or validating model assumptions. Furthermore, these same methodologies will be applied to assess the value of additional data, making it possible to quantitatively assess the potential of experimental upgrades or additional statistics. Students and postdocs will be trained in an emerging science centered around the analysis of complex systems. In addition to their own scientific discipline, they will be exposed to 21st century tools in statistics, computation and visualization.
