Figure 7.

Minimum first inference information out of one million independent experiments. This particularly rare case shows how first samples can mislead inference, which is later corrected by additional observations. The fourth inference (right) bears remarkably little overlap with the first.