What's the point of generating and testing a statistical model as opposed to simply performing experiments and making observations of wildlife populations to infer their dynamics? In what situation?