How Does the Error from Sampling to Big Data Change?

Abstract

In this article the authors aim to present a series of considerations, regarding the research carried out in the last 8 years, which starting from Big Data have posed different methodological problems related on the one hand to sampling and on the other to the conception of error in the scientific field. More precisely, the contribution will be divided into two macro areas of discussion. In the first part we will discuss sampling, with particular attention to break-offs and drop-outs and the relative response and cooperation rates, in order to understand how much these rates can still be valid in web 2.0 contexts. But at the same time we should ask whether it still makes sense to speak of probability sampling when in the hard sciences only a few cases are used in experiments, often less than a hundred. Further reflections concern the determination of a statistical representativeness which, especially online, can sometimes be overcome by an effective sociological representativeness. The second part of the contribution will be devoted to the discussion regarding biases and how the error can bring a series of further complexities in a pandemic reality. In this regard, the authors are convinced that an interpretative turning point must be made in the discussion that takes place around the error considered in the “science of discovery”.