Comparisons of p-adaptation strategies based on truncation- and discretisation-errors for high order discontinuous Galerkin methods