|
Abstract:
|
Many forensic sciences allege that the results of their comparisons yield uniqueness of source attribution. This uniqueness comes with certain assumptions that affect the resolution of their significance. The first is that all things are unique in space and, thus, their properties are nonoverlapping. The second is that properties are constant over time. These assumptions are not provable---the population size of "all things that might be evidence" is simply too large to account. Forensic science is therefore relegated to using statistics to interpret its results, but many forensic methods explicitly deny the applicability of statistics, such as fingerprints, while others deeply embrace statistics, such as DNA. This paper lays the foundation for why statistics should be used---in the proper evidentiary context---for all forensic analyses.
|
