Bayesian analysis with limited communication (Q1176754)

The \(i\)th individual in a group of \(m\) observes the random quantity \(X_ i\), where \(\underset\tilde {} X=(X_ 1,X_ 2,\dots,X_ m)\) has density \(g(\underset\tilde {} x\mid\theta)\) and \(\theta\) is an unknown parameter. Only a limited amount of information can be communicated and so the \(i\)th individual reports \(y_ i=h(x_ i)\). It is required to make inferences about \(\theta\) given a prior distribution on \(\theta\) and \(\underset\tilde {} y=(y_ 1,y_ 2,\dots,y_ m)\). Two versions of the problem are described. The first is where there is a common prior and \(y_ i\) is the posterior distribution of member \(i\) and the second is where \(\underset\tilde {} y\) is chosen to optimise the information communicated subject to a constraint on the amount that can be communicated.