Conner wrote:
I am trying to generate random numbers in a way that can be
replicated later (I am drawing random samples), and having
some trouble. I have two steps in my program. First, I
collapse the data set by group and randomly select groups
using the following code:
collapse $numvarssector, by($cluster)
set seed 54234
generate randomc=uniform()
sort randomc
g id2=_n
g ctreat=.
replace ctreat=1 if id2<=13
replace ctreat=0 if id2>13
sort cluster
drop randomc
I have no problem replicating the results of this step.
Next, after merging the grouped data with the individual
data, so that each person now has a variable "ctreat" which
indicates whether or not he or she is in a treatment group,
I draw random samples from within each treatment group.
What I want to do is the following: order each observation
within the groups by the size of a continuous variable,
"farm." Then, I want to break the observations into nearest
neighbor pairs using this variable, and randomly assign one
person from each pair to the treatment group:
sort cluster farm
by cluster: g id3=_n
sort $cluster id3
g invite=.
levelsof cluster, local(ctlevels)
foreach i of local ctlevels {
qui su id3 if cluster==`i'
forvalues j=1(2)`r(max)' {
set seed 54234
qui generate random`j'`i'=uniform() if
id3==`j'&$cluster==`i'|id3==(`j'+1)&$cluster==`i'
sort random`j'`i'
drop random`j'`i'
qui g d`j'_`i'=_n if
id3==`j'&cluster==`i'|id3==(`j'+1)&cluster==`i'
qui replace invite=1 if d`j'_`i'==1
qui replace invite=0 if d`j'_`i'==2
drop d`j'_`i'
}
}
I get a different sample every time I do this. Note that I
have already dropped the groups with ctreat==0 before doing
this step. Am I setting the seed in the wrong spot? Or
perhaps there is a better way to do this that I am not
seeing...
...
If I can follow up my own question, I looked harder at my
results, and the loop is generating the same two random
numbers for each pair, but it is switching who received
which number within each pair. Is there anyway I can keep
this from happening?
=========================================================
See:
help sort
and try:
sort cluster farm , stable
I believe it will solve your problem.
Svend
________________________________________________________
Svend Juul
Institut for Folkesundhed, Afdeling for Epidemiologi
(Institute of Public Health, Department of Epidemiology)
Vennelyst Boulevard 6
DK-8000 Aarhus C, Denmark
Phone, work: +45 8942 6090
Phone, home: +45 8693 7796
Fax: +45 8613 1580
E-mail: [email protected]
_________________________________________________________
*
* For searches and help try:
* http://www.stata.com/support/faqs/res/findit.html
* http://www.stata.com/support/statalist/faq
* http://www.ats.ucla.edu/stat/stata/
