Represents the randomization procedure Accelerated Biased Coin Design.
Usage
abcdPar(N, a, groups = LETTERS[1:2])
Arguments
N
integer for the total sample size of the trial.
a
nonnegative parameter which my be adjusted according to how strongly it is
desired to balance the experiment. a = 0 gives the complete randomization, while the
assignments become more deterministic as a increases.
groups
character vector of labels for the different treatments.
Details
This is a class of 'biased coins' where the probability of selecting the under-represented
treatment is dependent from the absolute difference between the two treatment allocations
up to the current step.
Value
S4 object of the class abcdPar.
References
A. B. Antognini and A. Giovagnoli (2004) A new 'biased coin design' for the sequential
allocation of two treatments.
Journal of the Royal Statistical Society. Series C (Applied Statistics)53, No. 4, 651-664