ABSTRACT
One must be careful, however, not to press the analogy between data analysis and criminal justice too far. Ordinarily, the phases of the criminal justice process are more or less sequential, and once the case has gone before the judge and jury, the investigative phase ends. With data analysis, particularly when dealing with spatial data, the process is more £exible. After an initial exploratory phase, the exploratory and con–rmatory processes to some extent co-occur, with results of the con–rmatory phase often motivating further exploration. The main danger that must be avoided is that of “data snooping,” that is, of letting the con–rmatory process be in£uenced inappropriately by the exploratory process through an introduced bias. Such an inappropriate in£uence would occur, for example, if the investigator selected for a hypothesis test subsets of the data that appeared in the exploratory phase to satisfy the hypothesis.
