Ordered sets are fairly new to PostgreSQL. The difference between an ordered set and a normal aggregate is that, in the case of an ordered set, the way data is fed to the aggregate does make a difference. Suppose you want to find a trend in your data—the order of the data is relevant.
Here is a simple example of calculating a median value:
test=# SELECT id % 2,
percentile_disc(0.5) WITHIN GROUP (ORDER BY id)
FROM generate_series(1, 123) AS id
GROUP BY 1;
?column? | percentile_disc
----------+-----------------
0 | 62
1 | 61
(2 rows)
The median can only be determined if there is sorted input.