FRT: Runs a Conditional Fisher Randomization Test
In gwb/RGroupFormation: Randomization tests for group-formation experiments

This function tests the null that H_0: Y(k) = Y(l) where 'k' and 'l' are the exposures of interest. The test statistic 'T' should be inputable (see description of argument).

1	FRT(Zobs, Yobs, T, A, k, l, n.rand = 1000, two.sided = FALSE)

`Zobs`	A list of numeric vectors. A group-assignment structure.
`Yobs`	A numeric vector of observed outcomes.
`T`	A function representing the test statistic. This function is called as followed in the code: 'T(W, Yobs.u, k, l, A.u)' where W, Yobs.u and A are the subsets of the exposure vector, observed outcomes vector and attribute vectors (respectively) for the focal units. The statistics 'T.dim' and 'T.studentized' can be used out of the box.
`A`	A numeric vector of attributes.
`k`	An integer representing an exposure value.
`l`	An integer representing an exposure value.
`n.rand`	An integer. The number of draws to use to build the (conditional) null distribution of the test statistic (defaults to 1000).
`two.sided`	A boolean. Whether to report a one-sided or two-sided p-value (defaults to FALSE).

# setup
N <- 90 # number of units
M <- 3 # number of units in each (equal-sized) group
A <- rep(c(0,1), each = N/2) # vector of binary attributes
L0 <- find.good.L0(N, M, 1000)

# generating a schedule of potential outcomes
# (for which H0^{0,1} is true, but H0^{0,2} is not).
Y0 <- rnorm(N)
science = cbind(Y0, Y0, Y0+1)
colnames(science) <- NULL

# Observing the assignments and outcomes
L <- rsL(L0, A, 1000)
Zobs <- Z.fn(L)
Yobs <- observe.outcomes(Zobs, A, science)

# running the tests
FRT(Zobs, Yobs, T.dim, A, 1, 0)
FRT(Zobs, Yobs, T.dim, A, 2, 0)