Could look for more data that supports the presupposed conclusion, then swap out the statistics? /s