Theoretical issues in PDF determination and associated uncertainties

We study several sources of theoretical uncertainty in the determination of parton distributions (PDFs) which may affect current PDF sets used for precision physics at the Large Hadron Collider, and explain discrepancies between them. We consider in particular the use of fixed-flavor versus variable-flavor number renormalization schemes, higher twist corrections, and nuclear corrections. We perform our study in the framework of the NNPDF2.3 global PDF determination, by quantifying in each case the impact of different theoretical assumptions on the output PDFs. We also study in each case the implications for benchmark cross sections at the LHC. We find that the impact in a global fit of a fixed-flavor number scheme is substantial, the impact of higher twists is negligible, and the impact of nuclear corrections is moderate and circumscribed.