Sample all people in a ped file. So the algorithm should generate a child out of two (sampled) parents. Maybe we have to implement a phased/non-phased model, because we can produces phased data with this approach!