Testing random

[Jussi Piitulainen]

sohcahtoa82 at gmail.com writes:

[...]

> Here's where I think you two are having such a huge disagreement.
> Does order matter?  It depends what you're pulling random numbers out
> for.
>
> The odds of seeing 2 1 are also only 1/36.  But if order doesn't
> matter in your application, then 1 2 is equivalent.  The odds of
> getting 2 1 OR 1 2 is 2/36, or 1/18.

[...]

I'm not sure what Thomas 'PointedEars' Lahn is talking about. It seems
to be something else than what others have been discussing.

Others have been discussing a record of the number of times that each
possible outcome came up in a sequence of random numbers. There is no
other record of the sequence. The number of drawings is much larger than
the number of possible outcomes. The subject line refers to testing
whether the record of counts is compatible with the drawings being
random in the usual sense: independent, with uniform distribution.

Someone pointed out that some numbers may not have occurred at all - I
think a piece of code needed modification - and so people have commented
on the probability of this happening ... and whether it depends on the
number of drawings.