It is interesting to see how posterior distributions are a compromise between prior knowledge and the likelihood. An extreme, funny case is a coin that is thought to be bimodal, say at 0.25 and 0.75. But when data was collected, it gave almost equal heads and tails.
