why does a sample census with plots located along a transect often provide more accurate data than a sample with plots that are located randomly?