genprob: Estimates probabilities of treatment assignment
In ri: ri: R package for performing randomization-based inference for experiments

Description Usage Arguments Details Value Author(s) References See Also Examples

Takes a permutation matrix and estimates the probabilities of treatment assignment for each unit

1	genprob(perms)

perms

N-by-k permutation matrix as produced by genperms or genperms.custom.

genprob is NOT intended to be used for complete randomization of clusters within blocks – instead, it takes an arbitrary permutation matrix and computes the proportions of random assignments for which each unit is in treatment. For simpler designs, genpermsexact should be used.

N-length numeric vector of values within the (0,1) interval, probability of treatment assignment

Peter M. Aronow <peter.aronow@yale.edu>; Cyrus Samii <cds2083@nyu.edu>

Gerber, Alan S. and Donald P. Green. 2012. Field Experiments: Design, Analysis, and Interpretation. New York: W.W. Norton.

genprobexact

## Rejected randomization scheme: reject if and only if there is significant imbalance

X <- c(1:200)

randfun <- function() {
  teststat <- -1
  while (teststat < 0.05) {
		Zri <- sample(c(rep(0,180),rep(1,20))) # imbalanced design
		fstat <- summary(lm(Zri~X))$fstatistic
		teststat <- pf(fstat[1],fstat[2],fstat[3],lower.tail=FALSE)  # extract F-test p-value
			}
	return(Zri)
}
perms <- genperms.custom(numiter=10000, randfun=randfun) # generate permutations
probs <- genprob(perms) # generate approximate probabilities from permutation matrix
cor(probs,(X-mean(X))^2) # observations with extreme X are less likely to be treated