Returns the six representations of Bowers and Burton, "A More Powerful Test for Randomized Experiments" (working paper), as a named list of functions, each mapping a block's outcome vector to a per-unit score vector of the same length:
- raw
the outcome itself (identity)
- rank
within-block mid-ranks,
riposte_rank()- mean_dist
mean L1 distance,
riposte_mean_dist()- mean_rank_dist
mean L1 distance on ranks,
riposte_mean_rank_dist()- max_dist
max distance to the block extremes,
riposte_max_dist()- huber
Huber's psi of the MAD-standardised outcome,
riposte_huber()
Details
Pass your own list (or this one extended) to riposte_test() or
riposte_components() to combine a different set. Each element must be a
function of one numeric vector returning a numeric vector of the same length;
it is applied within each block.
See also
riposte_reps_proposal() for the earlier set with raw tanh(y).