List Commands Library for 50g

[ttw]

(03-28-2018 08:12 PM)John Keith Wrote:  

(03-28-2018 01:43 PM)DavidM Wrote:  I like your method better, but I'd be less concerned about the cycle-length in this case than the issue of 0 having twice the representation of all of the other possible digits. I'm not familiar with why the command was created in the first place, though. Is there some mathematical reason that someone needing to generate a random matrix of single-digit integers needs more 0s than other digits? RANM has been around at least since the 48GX, but I have no idea what its origins are.

I would bet that there is a valid reason, although exactly what that reason is may be above my pay grade. Zeros do seem to predominate in "real-world" matrices, and I doubt that HP's programmers would make such a simple mistake.

This is just an intuitive guess, I have no evidence that this is actually the case. Also, the method suggested by ttw would be preferred for uses other than generating random matrix elements.

John

There are (as far as I know) no distributions with exactly double frequency between 0 and 1 from that everywhere else. Zero is no singled out at double the frequency. Normally, one might have a decrease in frequency from 0 to whatever. However, I have seen the problem I described create this distribution. I've done it (but caught the problem when reading the first test case) and I've seen other people do this. There's always a problem mapping one set of intervals into another; one has to be rather careful. One (not so obvious, it seems) problem is that a uniform distribution from 0 to 10 actually covers 11 different intervals. (Or 10 or 9 depending on the possibility of 0 or 10.)