Memorandum, Department of Economics, University of Oslo 2003,30
In this note I bring a new aspect into the so called Island Problem. Given that only cases where there is a suspect reaches the court, what is the consequence for the probability of guilt? I find that it indeed matters for the results that court cases are selected in this way. The analysis illustrates the general point that the exact protocol by which data are generated is an essential part of the information that should be used when analyzing data.