In article <[EMAIL PROTECTED]>,
David Heiser <[EMAIL PROTECTED]> wrote:

>"CybercafeUser" <[EMAIL PROTECTED]> wrote in message
>news:[EMAIL PROTECTED]
>> Excel's help mentions the Abramovitz (1969?) which you can also find as a
>> ref in 25 years old HP calculators.
>> It is not enough for most applications. The one in the addind for
>> statistical analysis seems to be different but Microsoft never answered my
>> questions;
>> Yves


>Abramowitz is sort of a icon. Every time you see this platinum icon, it
>refers to the classic publication of the old National Bureau of Standards in
>1963. Dover republishes it, so it is still in print. It is the gold standard
>on nummerical algorithms, and all kinds of everything else.  My copy is dog
>chewed (one corner) and just about worn out.

Abramowitz has some useful tables, but its approximations, etc.,
are quite dated and not all that useful.  It is an old icon.

>Excel used RAND as a worksheet cell function. It is a very simple one liner,
>very fast, and gives in most situations an acceptable result. Microsoft gave
>out the algorithm. There is a lot of RNG pedelling out there, and even more
>opinions on RAND.

I believe you will find almost all of them bad.

>The random number generator in the data analysis toolpak is different, one
>that was not invented by Microsoft, and Microsoft apparently cannot divulge
>it. It appears to be a simple congruential generator on a 16 bit integer.
>Has a short period. In general, don't use it. Dates back to the 1980's when
>anything programed for a computer was automatically accepted as being "the
>cat's meow".

If it is as you have stated, anyone could crack it.  Even long
period generators can be rather bad; I personally have a great
mistrust for pseudo-random number generators.




-- 
This address is for information only.  I do not claim that these views
are those of the Statistics Department or of Purdue University.
Herman Rubin, Department of Statistics, Purdue University
[EMAIL PROTECTED]         Phone: (765)494-6054   FAX: (765)494-0558
.
.
=================================================================
Instructions for joining and leaving this list, remarks about the
problem of INAPPROPRIATE MESSAGES, and archives are available at:
.                  http://jse.stat.ncsu.edu/                    .
=================================================================

Reply via email to