Abstract

The purpose of this article is to carry out a critical analysis of the role played by statistical inference in quantitative social research, with special reference to the Italian case. First, it is argued that many Italian sociologists tend to neglect the uncertainty that inevitably surrounds their research results, with the consequence that the validity of their conclusions if often put into question. Second, it is shown that among those who do care about statistical inference, the prevailing approach is the null hypothesis significance testing; the general inadequacy of this approach is demonstrated and its abandonment is invoked as necessary. Finally, several tools for measuring the uncertainty that surrounds the estimates of interest (classical confidence intervals and bayesian credible intervals) are illustrated and discussed; it is argued that these tools should be used routinely by all sociologists engaged in quantitative social research.