Randomisation of training course for 30 sales staff
The two diagrams on this page show two different ways of randomly allocating treatments to the experimental units.
In the first method, random indices are generated (0 to 29) to decide on which unit gets each treatment in turn. As allocation of treatments progresses, it is often found that the selected unit already has a treatment, slowing the process.
The second method works through each unit in turn, randomly allocating treatments. The complication is that the probabilities must be revised each time a treatment is used, reflecting the number of treatments of each kind that remain to be allocated.
The best method uses two spreadsheet columns, with random numbers in the second.
Then sort the two columns in increasing order of column B. This gives a random permutation of the indices of the units.