Performance of likelihood ratio methods
In previous chapters, methods for the computation of likelihood ratios in forensic science for the statistical evaluation of evidence have been given. However, as has been seen in previous examples, the computation of LR values still remains a challenge. There are many reasons for this challenge, among them:
Moreover, if the conditions are extremely degrading, it may be that the magnitudes of LR values supporting the wrong proposition are large, leading to what is known as strong misleading evidence (Royall 1997), an effect that is highly undesirable.
It is essential for the interpretation of evidence in casework that the LR model performs well. Misleading LR values in court may lead fact finders to the wrong decisions. This idea is the main motivation behind the establishment of validation procedures for evidence evaluation methods. Validation procedures enable the control of procedures and hence ease the use of LR methods in casework. If the LR generated by a given method performs badly, then it should not be used in casework. If it performs well it should be used. Therefore, the validation of evidence evaluation methods should be based on a careful process of performance measurement.
The performance of analytical ...