A couple of weeks ago we finished a chapter for the forthcoming APA handbook of forensic psychology.  The following paragraph is jettisoned from the final version (those co-authors!) but I still think it’s an interesting observation about the homology assumption (admittedly of only marginal relevance to the good stranger!):

As it happens, geographic profiling is a unique example of the homology assumption because evidence suggests that it does not hold in all cases, but that it holds in a sufficient number of cases to be useful. Specifically, research has shown that one cannot differentiate the actions (i.e., pattern of offense locations) of offenders who live locally from the actions of those who commute into the area to offend (Canter & Larkin, 1993). Thus, the single offending pattern does not have a homogenous relationship with the characteristic of where the offender is likely to live. However, assuming that an offender is not a commuter, in contexts where over 80% of offenders commit crimes locally, enables correct inferences about home location in the majority of cases.