Re: st: looking for more efficient programming for randomly shuffling list of numbers

[Jeph Herrin <stata@spandrel.net>]

A more substantial point is that my solution doesn't generate
a list of 20 distinct integers - there can be repetitions in the
list. But here's a solution using -rowranks-

**************
ssc install rowranks
**************
  clear all
  set obs 9600
  set seed 85999283
  gen str60 sticker=""
  forv i=1/20 {
	gen num`i'=runiform()
  }
  rowranks num1-num20, gen(rank1-rank20)
  forv i=1/20 {
	replace sticker=sticker+" "+string(rank`i')
  }
  drop num1-num20 rank1-rank20
**************


On 8/27/2010 4:46 AM, Nick Cox wrote:
> A very minor point, but -ceil(20* runiform())- is also a nice way to
> get uniform integers 1 ... 20 (for "any value of 20").
>
> Yet another posthumous nod to Kenneth E. Iverson, who came up with
> the names floor and ceiling for the round up and down functions
> yielding integer results.
>
> Nick n.j.cox@durham.ac.uk
>
> Jeph Herrin
>
> How about
>
> ************** clear all set obs 9600 set seed 85999283 gen str60
> sticker="" gen num=. forv i=1/20 { replace num=floor(20*uniform())+1
> replace sticker=sticker+" "+string(num) } drop num **************
>
> ?
>
> On 8/26/2010 11:11 AM, Evelyn Ersanilli wrote:
>
> > For a survey I need to make stickers that can be used to randomly
> > select respondents within a household. These stickers should
> > contain the numbers 1-20 in a random order (most households will
> > have less than 20 members). Each sticker should have its own random
> > order, though it is not a problem if by chance some stickers are
> > the same. The stickers should be generated in such a way that each
> > order of numbers 1-20 should be possible and equally likely.
> >
> > I have written a do-file for this (see below), but it's pretty
> > inefficient. My first step was to make a data file with one
> > variable that has the numbers 1-20 (so 20 cases). Then I let the
> > do-file run to create 70 random lists of the numbers 1-20. The main
> > inefficiency is that I had to create a separate .dta file for each
> > sticker and then merge them. This is not so bad with 70 lists, but
> > with the 9.600 that I will need to create it is pretty annoying.
> >
> > The recent discussion about need for setting the random seeds also
> > got me a bit worried, but if I understood it correctly it
> > shouldn't affect my problem too much.
> >
> > All suggestions are welcome.
> >
> > Kind regards,
> >
> > Evelyn
> >
> > ****************************** local i = 1 while `i'<   71 {
> > //generates 70 vars with random numbers gen u`i' = runiform() local
> > i = `i' + 1 }
> >
> > foreach order of varlist u1-u70 { sort `order' gen
> > order`order'=var1 preserve keep order`order' gen n=_n save
> > "C:\Users\sticker source`order'.dta" restore }
> >
> > clear gen n=_n foreach order in u1 u2 u3 u4 u5 u6 u7 u8 u9 u10 u11
> > u12 u13 u14 u15 u16 u17 u18 u19 u20 u21 u22 u23 u24 u25 u26 u27
> > u28 /// u29 u30 u31 u32 u33 u34 u35 u36 u37 u38 u39 u40 u41 u42 u43
> > u44 u45 u46 u47 u48 u49 u50 u51 u52 u53 u54 u55 u56 u57 u58 u59 u60
> > /// u61 u62 u63 u64 u65 u66 u67 u68 u69 u70{ merge 1:1 _n using
> > "C:\Users\sticker source`order'.dta", noreport nogenerate } drop n
> > save "C:\Users\sticker pilot 1 marok.dta", replace xpose, clear
> > outsheet using "C:\Users\sticker pilot 1 marok - transposed.out",
> > replace ***************************************
>
> * *   For searches and help try: *
> http://www.stata.com/help.cgi?search *
> http://www.stata.com/support/statalist/faq *
> http://www.ats.ucla.edu/stat/stata/
*
*   For searches and help try:
*   http://www.stata.com/help.cgi?search
*   http://www.stata.com/support/statalist/faq
*   http://www.ats.ucla.edu/stat/stata/