(25C) Sampling without repetition

[Dieter]

(10-27-2018 03:52 PM)Thomas Klemm Wrote:  Here's a translation of your program for the HP-42S:

Code:

...
15 DSE 03
16 GTO 01
17 GTO 02
18▸LBL 01
...

One more case where something like INV DSZ would be useful (greetings to the TI community ;-)). This can be accomplished quite easily by having the DSE followed by a test that always tests false. Since X holds a positive number at this point, a simple x<0? will do.

Code:

...
15 DSE 03
16 X<0?
17 GTO 02
...

Slightly shorter, slightly faster, saves one label.
Edit: OK, in this special case it's neither shorter nor label-saving, simply because LBL 01 is required anyway. #-)

(10-27-2018 03:52 PM)Thomas Klemm Wrote:  

(04-27-2017 04:07 AM)nsg Wrote:  I wonder if it is possible to improve on the algorithm and generate a sample of modest k for arbitrary N, for example, a sample of 30 out of 10000000.

In this case we might simply generate them at random and hope there's no collision.

In such a case there is no significant difference between sampling with or without repetition. So it doesn't matter much if a number is drawn twice. Otherwise the idea is: record the sample in a (sorted) list, generate the next number, see if it's part of the list. If yes: get a new number, if not: insert it into the list. Hmmm... somehow this sounds like a job for the HP50. ;-)

Dieter