Jeph suggests the ceil() solution in a follow-up with the caveat "as
long as uniform() never returns precisely 0, which I've
never seen" --uniform() is 0 with probability about 1 in 4 billion, so
to be safe you could
gen y = x[ceil(_N*(1-uniform()))]
On Thu, Feb 26, 2009 at 2:33 PM, Austin Nichols <[email protected]> wrote:
With replacement:
gen y = x[round(_N*uniform(),1)+1]
should be
gen y = x[round(_N*uniform()+.5)]
or
gen y = x[ceil(_N*uniform())]
I think.... (otherwise you can get _N+1 in the square brackets, for a
missing y, and the prob of getting y[1] is too low)
On Thu, Feb 26, 2009 at 2:19 PM, Jeph Herrin <[email protected]> wrote:
It's not clear if you want sampling with or without replacement.
With replacement:
gen y = x[round(_N*uniform(),1)+1]
without replacement:
egen index=rank(uniform())
gen y = x[index]
hth,
Jeph
*
* 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/
