Thanks, Peter.

The bucketizing I'm thinking of comes from
the "standard" way of selecting, say, integer values in
a range.

low-value + floor((high-value - low-value) * rand())

So if you have the pips on a die, then 0-.1666... become
1, etc.

I would think that at any given point in the generated
string of digits, the distribution might be skewed,
although it is uniform across the entire length before

Is there a desire to make the distribution uniform, no
matter where you are in it? Is there a measure of this?
An article on the Web mentioned a "super-duper"
generator as one that IBM has used. There were many
generators listed, with various metrics.
> Vernon,
> See inline comments...

> > I seem to recall problems with skewed distributions when generating
> > integers between 2 values. Does this "bucketizing" diminish the quality of
> > the distribution? The random number generator actually returns values
> > between 0 and 1.
> I think the bucketizing is the result of a poor seed.  Bucketizing would
> certainly diminish the quality of the distribution.
> _______________________________________________

This thread ...

Return to Archive home page | Return to MIDRANGE.COM home page