Literature Index

An attractive way of introducing Bayesian thinking is through a discrete model approach<br>where the parameter is assigned a discrete prior. Two generic R functions are introduced for<br>implementing posterior and predictive calculations for arbitrary choices of prior and sampling<br>densities. Several examples illustrate the usefulness of these functions in summarizing the<br>posterior distributions for one and two parameter problems and for comparing models by the use<br>of Bayes factors

Recently, Nathan (1986) criticized Bar-Hillel and Falk's (1982) analysis of some classical probability puzzles on the grounds that they wrongheadedly applied mathematics to the solving of problems suffering from "ambiguous informalities". Nathan's prescription for solving such problems boils down to assuring in advance that they are uniquely and formally soluble--though he says little about how this is to be done. Unfortunately, in real life problems seldom show concern as to whether their naturally occurring formulation is or is not ambigous, does or does not allow for unique formalization, etc. One step towards dealing with such problems intelligently is to recognize certain common cognitive pitfalls to which solvers seem vulnerable. This is discussed in the context of some examples, along with some empirical results.