[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]

RE: Re: st: random split

From   "Glenn Goldsmith" <>
To   <>
Subject   RE: Re: st: random split
Date   Sun, 12 Apr 2009 22:38:17 +0100


If you want exactly 267 in the first group, won't the following achieve this

  gen u = runiform()
  sort u
  gen split = _n <= 267

More generally (i.e. with _N!=800), you could use -ceil(_N/3)- or
-floor(_N/3)- to get 1/3, rounded up or down as desired.



"Martin Weiss" <> wrote:


"exactly a third of observations in the desired group"

How do you know that the split will be exact. Even leaving aside the
question whether ".33" is close enough to one third to warrant the
statement, the uniform number generator will still be noisy, i.e. the true
proportion will most likely be different from 1/3...


*   For searches and help try:

© Copyright 1996–2017 StataCorp LLC   |   Terms of use   |   Privacy   |   Contact us   |   What's new   |   Site index