Simulation screening and error control for big data.

"Big data + simulation model + post-simulation analysis" is viewed as a new research paradigm in the context of big data. To investigate an appropriate post-simulation analysis tool for the new paradigm, this paper proposes a new screening methodology based on using elementary effects, Bootstrap hypotheses testing, together with false discovery rate error techniques (abbreviated BFEE). BFEE is capable of precisely identifying and mining the most important factors among a large number of factors in simulated systems. Compared to the conventional elementary-effects-based method, BFEE guarantees accuracy and efficacy of simulation experiments. Moreover, BFEE exhibits higher flexibility and is easily adapted to the big data context in comparison with sequential bifurcation, an alternative screening method in simulation. The comparative analysis of Monte Carlo simulations indicate that the proposed method achieves desired simulation efforts and statistical precision, without the need for overmuch assumptions. The corresponding results from a real-world case study shows the potential applications in practice. [ABSTRACT FROM AUTHOR]