R Functions

Randomization

Download Randomization.R (GitHub)

• block_rand

  block_rand() is used for block randomization

  Inputs

  • n Total number of assignments needed. This should be (much) larger than the actual sample size.
  • groups Treatment group names, e.g., c('Control','Treatment')
  • prob Probabilities for block sizes. Does not need to sum to one (will be normalized).
    • If there are only two treatment groups, possible block sizes are 2, 4, or 6, with default probabilities of c(0.2, 0.7, 0.1).
  • allocation Allocation ratio. e.g., c(2,1,1) for a 2:1:1 allocation ratio. Use NA for equal allocation.
  • seed Random seed for reproducibility (Use NA for a random random seed)

  Example

  set.seed(620)
  out <- block_rand(n=40, groups=c('X','Z'), prob=c(2,7,1),
                    allocation=c(2,1), seed=NA)
  
  > out
  $seed
  [1] 1760
  
  $block_sizes
  [1] 3 6 6 9 6 9 6
  
  $assignments
   [1] "Z" "X" "X" "Z" "X" "X" "Z" "X" "X" "X" "X" "X" "X" "Z"
  [15] "Z" "X" "X" "X" "Z" "Z" "X" "Z" "X" "X" "Z" "X" "Z" "X"
  [29] "X" "X" "X" "X" "X" "X" "X" "Z" "Z" "Z" "X" "X" "X" "X"
  [43] "Z" "Z" "X"
  
  table(out$assignments)
  
   X  Z 
  30 15 
  

  Setting seed=1760 allows the result to be reproduced.

  out <- block_rand(n=40, groups=c('X','Z'), prob=c(2,7,1),
                  allocation=c(2,1), seed=1760)
  

• strat_rand

  The strat_rand() function performs stratified randomization, utilizing block_rand() internally.

  Inputs

  Main Arguments

  • stratum_size Sample size per stratum (each subgroup receives this number of allocations).
  • groups Treatment group names; refer to block_rand().
  • prob Probabilities for block sizes; refer to block_rand().
  • allocation Allocation ratios; refer to block_rand().
  • seed Random seed for reproducibility; refer to block_rand().
  • ... Stratification variables as named vectors (e.g., Hospital=c('V','X','Y')).

  Output

  assignments contains S+1 columns, where S is the number of strata. group column contains the actual allocations, while the remaining columns identify each stratum. This is the format required by REDCap for its stratified randomization function.

  Example

  set.seed(620)
  out <- strat_rand(stratum_size=40, groups=c('X','Z'), prob=c(2,7,1), 
             allocation=c(2,1), seed=NA,
             Hospital=c('V','A','B'), Smoking=c('Yes','No'))
  
  > out$seed
  [1] 2032
  
  > out$assignments
     group Hospital Smoking
  1      X        V     Yes
  2      X        V     Yes
  3      Z        V     Yes
  4      X        V     Yes
  5      X        V     Yes
  6      Z        V     Yes
  7      Z        V     Yes
  8      X        V     Yes
  9      X        V     Yes
  10     X        V     Yes
  
  ...
  ...
  
  > out$strata
  [[1]]
  [[1]]$seed
  [1] 6245
  
  [[1]]$block_sizes
  [1] 6 3 6 6 6 6 6 6
  
  [[1]]$stratum
    Hospital Smoking
  1        V     Yes
  
  
  [[2]]
  [[2]]$seed
  [1] 3140
  
  [[2]]$block_sizes
  [1] 6 6 6 6 6 3 3 6
  
  [[2]]$stratum
    Hospital Smoking
  2        A     Yes
  
  ...
  ...