agreed, RANTBL is extremely slow when # of bins are large, say 100K.
as for the syntax perspective, it accepts 'of...' style syntax, so I
usually use array to avoid tedious list of probabilities:
array _p{*} p1p100;
do j=1 to 100; _p[j]=0.01; end;
seed=3454353;
x=rantbl(seed, of _p[*]);
On Thu, 16 Sep 2010 11:16:05 0700, Dale McLerran
<stringplayer_2@YAHOO.COM> wrote:
>The RANTBL function shines when the number of bins is small AND
>when the probability varies across bins. Here, where we are
>selecting an observation from among 100K bins, the RANTBL function
>would not be advisable. Moreover, since the bin probabilities
>follow a uniform distribution, it is much simpler to employ
>something like
>
>x = ceil(100000*ranuni(seed));
>
>
>For 10 bins with probabilities conforming to a uniform distribution,
>compare
>
>x = rantbl(seed,.1,.1,.1,.1,.1,.1,.1,.1,.1,.1);
>
>with
>
>x = ceil(10*ranuni(seed));
>
>I would certainly prefer the latter, even though the number of
>bins is not particularly large.
>
>Dale
>
>
>Dale McLerran
>Fred Hutchinson Cancer Research Center
>mailto: dmclerra@NO_SPAMfhcrc.org
>Ph: (206) 6672926
>Fax: (206) 6675977
>
>
>
> On Thu, 9/16/10, oloolo <dynamicpanel@YAHOO.COM> wrote:
>
>> From: oloolo <dynamicpanel@YAHOO.COM>
>> Subject: Re: how to generate 1 to 100k random numbers without zeros?
>> To: SASL@LISTSERV.UGA.EDU
>> Date: Thursday, September 16, 2010, 7:03 AM
>> all replies are wonderful
>> discussions, but I am a little surprised that no
>> body mentioned using RANTBL to generate the random variable
>> as the question
>> is a discrete r.v generation problem in nature
>>
>> On Mon, 13 Sep 2010 10:52:22 0700, Arthur Tabachneck
>> <art297@NETSCAPE.NET>
>> wrote:
>>
>> >Kate,
>> >
>> >Whether it is beneficial depends upon what you are
>> trying to do. If
>> >you are trying to select random numbers between 1 and
>> 100000, without
>> >replacement, then datanull's proposed code is the one
>> offering that
>> >accomplishes the task. Unless I overlooked
>> something it appeared that
>> >all of the other offerings would result in many of the
>> same numbers
>> >being picked repeatedly.
>> >
>> >HTH,
>> >Art
>> >
>> >On Sep 13, 8:01 am, kate <juanz...@gmail.com>
>> wrote:
>> >> thank you.
>> >> what's the benefit of this code instead of using
>> random_num =
>> >> int(ranuni(0)*99999)+1 ?
>> >>
>> >> thanks
>> >>
>> >> On Sep 13, 7:53 am, "data _null_;" <datan...@gmail.com>
>> wrote:
>> >>
>> >>
>> >>
>> >> > On Sep 13, 6:46 am, kate <juanz...@gmail.com>
>> wrote:
>> >>
>> >> > > interestingly, the very first time I
>> run
>> >> > > random_number=int(ranuni(1))*100000, I
>> got 2 zeros. which caused
>> >> > > unexpected mistake in final result.
>> >>
>> >> > > On Sep 13, 3:48 am, Patrick <patrick.mat...@gmx.ch>
>> wrote:
>> >>
>> >> > > > Barry
>> >>
>> >> > > > You will find many SAS examples
>> where ceil() is used as I
>> suggested.
>> >>
>> >> > > > I would assume that the likelyhood
>> to get a exact 0 from ranuni()
>> is
>> >> > > > neglectable (if the underlying
>> algorithm creates a 0 at all).
>> >>
>> >> > > > Cheers
>> >> > > > Patrick Hide quoted text 
>> >>
>> >> > >  Show quoted text 
>> >>
>> >> > PROC PLAN can produce a random permutation of
>> the values 1100000;
>> >> > there will be no zero.
>> >>
>> >> > proc plan seed=1091883594;
>> >> > factors r=100000 / noprint;
>> >> > output out=r;
>> >> > run;
>> >> > proc print data=r(obs=100);
>> >> > run;
>> >> > proc print data=r(where=(r=0));
>> >> > run; Hide quoted text 
>> >>
>> >> >  Show quoted text  Hide quoted text 
>> >>
>> >>  Show quoted text 
>>
