Third, and perhaps most important, there has been a dearth of efficient, easy to use, mainstream statistical software for Bayesian analysis.While these methods work very well when there is prior information, often in applied work there is little such information available. Second, Bayesian methods require a prior distribution to be specified for the parameters to be estimated.First, and possibly due to the next two reasons, applied statistics courses have mostly not spent much time on Bayesian approaches.There are, I believe, three main reasons for this. Applied statisticians, though, have until recently been overwhelmingly in the frequentist camp.
(But see also this link for a vigorous debate on this: What's wrong with this?). Googling Bayesian versus frequentist produces a vast collection of items on this topic. For decades, statistical theorists have debated the merits of the classical or frequentist approach versus the Bayesian approach.
0 kommentar(er)
