Blog

Epistemological vigilance and sampling

Some days ago I was browsing on my Twitter page, and I saw that one of my contacts had published a request in full good faith. A friend of his was conducting a study for the master course he was doing and the former asked his contacts to respond the online survey his friend uploaded to a website. I will leave aside the details about the object of study, but I will say that from a methodological point of view the study intended to “have a knowledge of the current situation of this sector [education]”. That is, the study sought to get representative data on issues related to training. The most appropriate sampling strategy to achieve this is to perform a random sampling. And therein lies the problem.

Asking for “volunteers”, who would be caught among our contacts, to respond the survey of our friend means roundly violating the criterion of randomness. Why? It is very simple: the contacts are chosen based on the affinities. That is, we tend to have friends who like the same things we like: it is clear that people are not chosen randomly: the first passing through the street is not being asked to “be our friend”. About this there is a lot of writing, and one of the best books on the subject, documenting without a doubt this statement is Distinction, a tour de force and a work of Pierre Bourdieu to which this social scientist devoted fifteen years of research. It is the second most important book of all time in the social sciences according to the ISA (International Sociological Association).

From a methodological point of view choosing a sample strategy such as asking our Facebook or Twitter contacts to respond a survey —which would be the equivalent to a snowball sampling—, when in fact it would require an strictly random sampling, implies that the sample obtained is not representative, which means, from the point of view of the results and conclusions obtained, that data are not reliable. That is, this represents a violation of the criteria of representativeness and reliability. We cannot rely on the findings of this study. And that only because we have choose the wrong way to attract the subjects of study, since we have not thought about it with the necessary rigor, as we have introduced a bias that is clearly documented in any manual method of sample surveys: the so-called selection bias.