On family-based genome-wide association studies with large pedigrees: observations and recommendations