Misinformation and Disinformation in Statistical Methodology for Social Sciences: causes, consequences, and remedies

The present manuscript is an introductory summary on the topic of misleading statistical inference: on the basis of the results of a peer-reviewed study that adopted a probabilistic methodology, a fact is claimed as scientific, but still, there are solid reasons for doubting about the veracity of this claim. For example, results from another scientific study are incompatible with those of the first. This is called absence of replicability in studies. The focus is on Social Sciences because here the Replication Crisis has been paradigmatic for a pledge to improve the methodological standards in scientific research. If the fabrication of a photograph of a new insect is a form of scientific disinformation that can be spotted by an expert entomologist only, issues such as the p-hacking in clinical studies on human behavior, or the sensitivity to the choice of time lag variable in a model of economic performance had involved wider scientific communities. Sometimes these studies become central elements of justification for decision-making policies. In this regard, it is noteworthy the commitment of the largest public research agency, the DARPA, to the SCORE Project for investigating features of scientific papers that are predictive of reliability of its claims. Many of the proposals that came out from the Replication Crisis, although not fully integrated as new standards, has been fecund attempt to make progresses in statistical methodology for Social Sciences. The diffusion of statistical software for Causal and Bayesian analysis, the practice of pre-registration of studies, and the re-kindled interest for the concept of model variance through new graphical methods (e.g. Specification Curve) are all reactions against an epistemological crisis that and are worth to be addressed by any researcher interested in modern Statistics.