Archives of the TeradataForum
Message Posted: Fri, 09 Jan 2004 @ 21:34:33 GMT
| Subj: || || Re: Sampling Method |
| From: || || Victor Sokovin |
| ||We currently have an auditor using TD to select row samples. I have some "opinions" on the way sampling works, but no authoritative
information. Given that we are currently on V2R3 and have no RANDOM function, does TD:|| |
| ||1. Move all rows into spool before selecting the sample?|| |
Not sure about the spool but at least in-memory.
| ||2. Use a.some internal random sampling algorithm or|| |
| ||b.simply select each nth record given the whole population/sample size ratio to create the sample from the spool?|| |
Surely not. Just try the same query with the sample clause n times.
| ||3. Can anyone identify the "published" technique TD uses for sample selection by name or formula?|| |
I recall Geoffrey Rommel has explained this once on this forum and I have managed to look up the thread. Just search for the thread entitled
"Random Sampling" in the Feb 2003 archive.