[R] Iterative sampling with restrictions

[Daniel Nordlund]

> -----Original Message-----
> From: r-help-bounces at r-project.org [mailto:r-help-bounces at r-project.org]
> On Behalf Of David A. Kim
> Sent: Monday, August 20, 2012 9:17 PM
> To: r-help at r-project.org
> Subject: [R] Iterative sampling with restrictions
> 
> Hi all,
> 
> I'm working on a seemingly trivial problem that I can't figure out how
> to implement in R.  I'd be most grateful for any help here.
> 
> I want to do the following: first, randomly assign each of n units to
> one of g groups of size n/g. Then, randomly re-assign each of the n
> units to a different group (i.e., same as the first step, but the unit
> can't be assigned to a group to which it's already belonged).  Then
> repeat this step until each unit has at some point been assigned to
> every group.
> 
> More concretely, say I have 60 units and 3 groups into which to divide
> the units.  I could first do something like:
> 
> group1<-sample(1:60,20)
> group2<-sample(setdiff(1:60,group1),20)
> group3<-sample(setdiff(1:60,c(group1,group2)),20)
> 
> But then how to randomly re-assign group membership such that all
> units are assured a different group assignment in the second "wave" of
> grouping?  Just narrowing the sampling pool to those units that
> weren't previously assigned to a given group won't work (consider the
> case where groups 1 and 2 swap units: in the second wave, there would
> be no units to assign to group 3 as all the remaining units had
> already been in group 3 in the first wave).
> 
> Most grateful for any assistance,
> 
> David
>

David,

I would collect the sample waves into a data.frame.  I am sure someone will be able to help you with a more general and/or efficient solution, but to get you started I have provided one possible solution to your 60 unit 3 wave example
 
#create data.frame with IDs
df <- data.frame(id=1:60)

#create first sample wave
df$wave1 <- sample(rep(1:3,20))

#reorder df and create second wave sample
df <- df[order(df$wave1),]
df$wave2 <-  c(sample(rep(c(2,3),10)),
               sample(rep(c(1,3),10)),sample(rep(c(1,2),10)))

#now use set diff to create 3rd wave
for(i in 1:60) df[i,'wave3'] <- unlist(setdiff(1:3,df[i,2:3]))

df


Hope this is helpful,

Dan

Daniel Nordlund
Bothell, WA USA