Calculate percentile in bash

percentile.sh

#!/bin/bash
# stdin should be integers, one per line.
percentile=$1
tmp="$(tempfile)"
total=$(sort -n | tee "$tmp" | wc -l)
# (n + 99) / 100 with integers is effectively ceil(n/100) with floats
count=$(((total * percentile + 99) / 100))
head -n $count "$tmp" | tail -n 1
rm "$tmp"

Great idea with (n + 99) / 100 👍

You may use "sed -n ${count}p $tmp" instead of "head | tail"

Made a one-liner version of this for fun. The trick is to count the number of lines and then select the right line based on that count within the same pipeline and without temp files:
cat mynumbers.txt|(percentile=95; (sort -n;echo)|nl -ba -v0|tac|(read count;cut=$(((count * percentile + 99) / 100)); tac|sed -n "${cut}s/.*\t//p"))