Randomisation, or “random allocation” is at the heart of the randomised controlled trial and a very strong way to remove bias from experiments. Combined with statistics random allocation to different intervention creates a very powerful way to make causal attributions: if something that looks highly unlikely to have happened randomly is seen, then it becomes persuasive to assume that something systematic is happening, something with a causal root.
There’s a funny quirk to this that true randomness is hard to achieve, few coins or dice are truly unbiased (though the level of bias may not matter realistically) and most of the time when people talk about having randomised something they will have used a “pseudo random number generator” (PRNG): one of a collection of mathematical/computational algorithms that produce a sequence of numbers whose sequence is largely indistinguishable from that which would be expected of a truly random process. Such PRNGs are not truly random and started at the same point will always produce the same sequence of numbers … which has the nice property that if one uses them correctly they give something which is as nearly random as we need, but is perfectly reproducible: the rather paradoxical sounding reproducible randomness.
Random processes are sometimes referred to as “stochastic“.
Try also #
Randomised Controlled Trial (RCT)
Causality & causal attributions
Mostly chapter 10.