BAYESIAN INFERENCE AS AN ALTERNATIVE FOR A BETTER COMPREHENSION OF UNCERTAINTY IN APPLIED STATISTICAL PROBLEMS

Abstract

In this talk the problem of inverse probability or Bayesian probability is addressed. The frequentist or classic point of view of statistics has dominated the vast realm of applications in the different fields where statistical analysis is required. This dominance is due to different reasons, and one of them is that frequentist inference can be easily performed by just following a simple recipe of steps (hypothesis testing). This seeming advantage of the frequentist point of view contrasts with the use of Bayesian inference, which requires the practitioner to have a basic understanding of probability theory and some computational skills. However, during the last decade or so, the community of Bayesian statistics has developed a significant number of easy-to-use software applications for Bayesian analysis and some new algorithms for approximation that have facilitated the use of Bayesian analysis by people with few mathematical and computational skills. Thus, the advantages of the Bayesian paradigm and how it provides a more appropriate way to measure uncertainty regarding scientific hypotheses are shown through different applications.