ABSTRACT
Vitriolic arguments about the merits of Bayesian versus classical approaches seem to have
faded into a quaint past of which current researchers in the social sciences are, for the most
part, blissfully unaware. In fact, it almost seems odd well into the 21st century that deep
philosophical conflicts dominated the last century on this issue. What happened? Bayesian
methods always had a natural underlying advantage because all unknown quantities are
treated probabilistically, and this is the way that statisticians and applied statisticians really
prefer to think. However, without the computational mechanisms that entered into the field
we were stuck with models that couldn’t be estimated, prior distributions (distributions
that describe what we know before the data analysis) that incorporated uncomfortable
assumptions, and an adherence to some bankrupt testing notions. Not surprisingly, what
changed all this was a dramatic increase in computational power and major advances in the
algorithms used on these machines. We now live in a world where there are very few model
limitations, other than perhaps our imaginations. We therefore live in world now where
researchers are for the most part comfortable specifying Bayesian and classical models as it
suits their purposes.