Builds a 0/1 assignment matrix whose first column is the observed treatment
and whose remaining nresample columns are independent re-randomizations,
each drawn by sample() within block so that every block keeps its observed
number of treated units. This is the randomization reference for the Monte
Carlo combinations.
Usage
riposte_block_draws(z, block, nresample = 1999L)
Arguments
- z
integer/numeric 0/1 vector of observed treatment assignment.
- block
factor or vector of block labels, the same length as z.
- nresample
number of re-randomizations (columns 2..nresample+1).
Value
a numeric matrix with length(z) rows and nresample + 1 columns;
column 1 is z. Row order matches the input.
Details
For a cluster-randomized design, pass cluster-level vectors (z and block
indexed by cluster, one entry per cluster); the permutation is then over
cluster assignments, which is the correct reference for clusters.