Blocking is a method of restricted randomisation that ensures the treatment groups are balanced at the end of every block. For example, here are two permuted blocks of 4 with treatment groups A and B:
A B B A B A B A
Random permuted blocks are blocks of different sizes, where the size of the next block is randomly chosen from the available block sizes. For example, here is a list of random permuted blocks of sizes 4 or 6:
A A B A B B A B A B B B A A B A A B A B A B B A B A A A B B
Blocking can be used within strata, so that important prognostic characteristics (the stratification factors) are balanced between the treatment groups:
Men | A B A B A A B B B A B B A B A A B A A B |
Women | B B A A B A A B B A B B A A A B B A |
Using this list the frequencies after 9 men have been recruited and 5 women will be:
A | B | Total | |
---|---|---|---|
Men | 4 | 5 | 9 |
Women | 2 | 3 | 5 |
Total | 6 | 8 | 14 |
Block sizes must be multiples of the number of treatments and take the allocation ratio into account. For 1:1 randomisation of 2 groups, blocks can be size 2, 4, 6 etc. For 1:1:1 randomisation of 3 groups or 2:1 randomisation of 2 groups, blocks can be size 3, 6, 9 etc.
The treatment allocation is predictable towards the end of a block. For this reason block sizes should be kept confidential and not shared with those randomising. Large blocks reduce predictability, but will not restrict the randomisation as closely as small blocks. If interim analyses are planned at particular sample sizes, it is desirable that the treatments are balanced at these points. Having many stratification factors can lead to many incomplete blocks and thereby imbalance. Therefore choice of block size(s) should take into account the sample size, planned interim analyses and number of stratification factors.
You can experiment with different block sizes and stratification factors on our simulation page. This will show you how much imbalance to expect for various choices.