> As an aside, thinking about this (and many other things) in terms of ordinal rankings will often lead you down the wrong path. "Top four reasons" tells us not

> As an aside, thinking about this (and many other things) in terms of ordinal rankings will often lead you down the wrong path. "Top four reasons" tells us nothing at all. If there are 5,000 categories constructed and they are roughly evenly distributed, "top four" tells us nothing. If 99% of the probability mass is in #1, "top four" tells us nothing. That is correct, but please note that it's still a significant component. In the 15-24 age group, the top 1 cause has 13.4k deaths, the top 2 has 5.7k, and homicide has 5.1k. That's not very far away in relative terms.