| Popis: |
When performing data analysis, a researcher often faces a choice between Frequentist and Bayesian approaches, each of which offers distinct principles and prescribed methods. Frequentism operates under the assumption of repeated sampling, aiming for so-called objective inferences through significance tests and efficient estimators. Bayesianism, on the other hand, integrates a researcher's prior beliefs about a hypothesis while updating these with new evidence to produce posterior distributions. Despite the technical rigour of both methods, neither approach appears universally applicable. A single, "correct" statistical school may seem like an objective ideal. However, we will see that it becomes impossible to choose between the two schools, even when we try our best to fulfil this ideal. Instead, this essay proposes a context-dependent approach to guide the selection of an appropriate statistical school. This approach style is not novel. Worsdale & Wright (2021) presents Douglas (2004)'s "operational" objectivity in the search for an objective gender inequality index. The authors point out the worrying obsession researchers have to find a single universal true measure of gender inequality. Rather, Worsdale & Wright (2021) recommend taking the research goals and context into "objectivity", making a context-dependent objectivity. I take the same idea and apply it to the search for a normative system of statistics: contextualizing statistical norms. |
