The fact that the same analysis can yield wildly different findings simply by changing which of the supposedly interchangeable datasets is used, reminds us just how much our results depend on the data we look at, rather than divining universal truths about our world.