Christian Hennig - Understanding statistical inference based on models that aren't true - Perspectives on Scientific Error 2024
For slides, see osf.io/ayfek/
Statistical inference is based on probability models, and most of the theory behind it assumes these models to be true. But models are idealisations, and it makes little sense to postulate that they are literally true in reality. Models are however required to analyse the behaviour of statistical methods in any generality. In order to explore the implications of running statistical inference based on models that aren't true, it is helpful to look at more general supermodels that allow for violation of the supposedly assumed models. I will present a framework for how to think about statistical inference based on models that aren't true, conditions under which such inference can be useful or misleading, and what impact this has on the interpretation of the results in practical settings.
Негізгі бет Ғылым және технология Christian Hennig - Understanding statistical inference based on models that aren't true - PoSE 2024
Пікірлер