It’s coming up to Christmas so I thought I would be a bit of a Grinch 🙂 As someone who has refereed a lot of papers in my time, one of my particular bugbears is when I come across the phrases, “taken at random”, “sampled randomly” or variations thereon. My edition of the OED defines at random as “haphazard without aim or purpose, or principle, heedlessly”; the statistical part of the definition qualifies this further as “equal chances for each item to be selected”. Whenever I see the word random in the methods and materials section I annotate the paper with the phrase “truly random or haphazardly?” Almost without exception*, when the author responds to my query, it is to admit that in reality they meant haphazardly.
There is a commonly held belief among field biologists that random sampling can be quickly and safely done by standing in a field and throwing a quadrat over their shoulder or closing their eyes and throwing the quadrat into the air. The late great Sir Richard Southwood deals with this myth in his usual no nonsense style “Biologists often use methods for random sampling that are less precise than the use of random numbers, such as throwing a stick or quadrat. Such methods are not strictly random” (Southwood, 1966). If you have ever tried this yourself, you will, I hope, be the first to admit, that you position yourself in all sorts of non-random ways, to make sure that the quadrat is not going to get lost, get hung-up in a tree, end up in a lake or river or miss the only green bit of vegetation in the field. Other so-called random approaches include the walking around the tree/into the meadow/along the path approach and examining the first leaf/branch/plant you come across after x number of steps and counting what you see on that. Again, this is equally subject to being confounded by the terrain and location of the site, and it is a rare person who isn’t subconsciously swayed for or against a leaf because of its appearance. I was convinced that this mode of sampling, which is more accurately described as haphazard, was commonly called professorial random sampling. A recent request by me on Twitter for people to tell me if they had heard of, or used the term themselves, resulted in a zero response rate, so perhaps it was just something we used in our lab. Of course, it wasn’t a random survey so I shouldn’t read too much into it 🙂
So, if you are going to claim that you sampled randomly or selected/arranged randomly, make sure you use a random number generator. It is very simple to do, although somewhat time-consuming to implement in reality. When I was a student, most good statistics books included among all the other useful tables, a page of random numbers to help you meet a state of true randomness.
Pre-prepared random numbers from my copy of Sokal & Rohlf (1973)
Nowadays, you can, if you use Excel, generate random numbers using the function RAND. Those of you who are not fans of Excel can try this handy link https://www.random.org/sequences/
If you’re reading this, you now have no excuses left. If you are going to claim that you did something randomly make sure you actually did so, or confess that you sampled haphazardly; it is nothing to be ashamed of 🙂 and is much faster than true random sampling, hence its popularity. Alternatively, you can avoid the whole issue and sample along a stratified transect or arrange your experimental blocks using a Latin Square.
Sokal, R.R. & Rohlf, F.J. (1973) Introduction to Biostatistsics. W.H. Freeman & Company, San Francisco.
Southwood, T.R.E. (1966) Ecological Methods. Chapman & Hall, London.
*I have, on a few occasions, had an author respond that yes, they did indeed use random number tables and/or generators.