Archives of the TeradataForum

Message Posted: Thu, 20 Feb 2003 @ 11:22:28 GMT

	<Prev	Next>		<<First	<Prev	Next>	Last>>

Hi Ray

I guess for a 'Truly Random' sample, you would need to connect your teradata to a truly random analogue device, like a roulette wheel or a lotto machine. Pedantics aside, the only facilities in v2r4 for this are RANDOM and SAMPLE. Random can always be statistically tested, (See Knuth - Seminumerical Algorithms), but this wouldn't answer any questions as to whether the method was sound. It would be nice to have a better description of how the sequence works, is seeded etc. in the documentation. As to the statistical tests, I have performed a few on modest sized runs on ours and found it to be adequate (i.e. it doesn't fail any chi-square frequency tests on samples of up to about size 10000 that I have tested it on)

Regards

John

	<Prev	Next>		<<First	<Prev	Next>	Last>>