ABSTRACT
The essential difference between the approach to inference which we have adopted until now and the alternative approach, called the Bayesian approach which does not shrink from using probabilities to specify degrees of belief, is that in the latter there is one further ingredient in a general mathematical model. The probability distribution describes our degrees of belief in possible parameter values prior to an observation being made, and consequently it is called a prior distribution. The Bayesian approach may be used to justify most intuitively acceptable parametric procedures and for an account of this the reader is referred to D. V. Lindley. The essential simplicity of Bayesian theory is attractive. It is not difficult to apply once a prior distribution has been assigned. But to assign one in practice may be an extremely difficult problem.
