In a blog post, Daniel Lemire complains that in research papers, people express a measure or a gain in absolute value, and then go to a conclusion about optimality due to the numbers obtained. Lemire calls this **fallacy of absolute numbers**. With a comment to this post, John Cook gives the correct term coined for this situation: **numerator-only data**. I’ve ever heard that phrase for the first time. I’ve found it interesting, and wanted to share it with a blog post.

So, what exactly is **numerator-only data**? Cook explains the term in a blog post as **data without anything to compare it to, no denominator**. It is the data that leaves us asking “compared to what?”.

For example, if one tells that an athlete runs 100 m sprint in 9.75 seconds, we need to ask “is it good or bad”. The number itself does not give enough information, it is nothing more than a data value. If he tells that the world record is 9.58 seconds, then we have a denominator to compare it to; that is, 9.75 seconds is now meaningful. Of course, we may already have the knowledge of the “denominator”, but this may not always be the case, and as Lemire points out that numerator-only data becomes a problem in research papers, for instance.