Speaker
Description
The production of useful and high-quality nuclear data requires experimental measurements with high precision and extensive information on uncertainties and possible correlations. When performing experimental measurement data analysis, the physicist uses many external parameters (detector efficiencies, distance of flight, …) in addition to the raw data. All the steps of the data processing (event selection, calibration, …) may introduce uncertainties and correlations.
The usual method for combining and computing uncertainties is to use analytical developments based on the perturbation theory (e.g. 〖u_(f(x))〗^2=(∂f/∂x)^2×u_x^2). This method works well for simple cases, but with multiple parameters and sources of uncertainty, deriving the final total combined uncertainty can be long and complex. Reporting the formula into the analysis code becomes a tedious process where mistakes can appear and the final uncertainty value will be wrong.
Furthermore, it strictly applies only to small deviations from the central values, which is not always the case. Finally, this method makes calculation of covariances hard and the inclusion of some unusual form of uncertainty (assymetric, non Gaussian) even more difficult.
To overcome this issue, one may rely on random sampling methods (a.k.a Monte Carlo). With modern computing infrastructures, large storage space and many computing units are available to process data. This allows, within a reasonable time and coding budget, the repeated processing of data with variation of input parameters within their nominal distribution probability. From the collection of results, a central value, uncertainties and covariances can be extracted easily.
The poster will present an example of how the method is used in production of (n, n’ gamma) cross-sections with uncertainties. The advantages and drawbacks will be discussed, and the method will be compared to the analytical one.