Table stats in 10.2

Platform - Oracle EE 10.2 on AIX 5.3

User complains that application runs fine on Monday but is slow by
Friday. We think we have found an issue with OS i/o (MetaLink note
271444.1 and related), but in getting there I took the usual route of
checking the currency of the table stats. Selecting dba_tables for on
OWNER and sorting on last_analyzed, we found that most of the tables
had been analyzed in the last six days, but they were in groups ...
some analyzed Apr 23, some Apr 24, etc. There was also a group of
tables, all name TEMP_something, that had nulls for LAST_ANALYZED.
Thinking the pattern of date/time of analysis rather odd, we went
looking for how it was being done.. how it was scheduled. Nothing in
crontab. Nothing in DBA_JOBS. All of the tables that have stats show
'YES' for MONITORING; most but not all show YES for GLOBAL_STATS and
those that don't constitute all of table in one of the time slots for
which stats were gathered. None of the TEMP_* tables have stats, and
all of them have both GLOBAL_STATS and MONITORING set to NO.

Is there an easy explanation to how these stats are being gathered
without some manually scheduled job? And which manual do I need to
reference to get up to speed on this?

Thanks.

.