Using hierarchical Bayesian modelling to assess shedder test suitability

Advances in trace DNA analysis, in combination with increased ability to detect minute amounts of DNA, necessitated research investment into the understanding of the factors relevant to DNA-TPPR, including individuals' ability of deposit their trace DNA, termed their shedder status. Variability in shedder tests and assessment methods, in combination with limited data routinely published, currently limits a scientist's ability to utilize such data in casework circumstances while also raising questions on which of the myriads of methods is the most appropriate going forward. In the present study, we investigated trace DNA deposition by 5 individuals gripping a plastic tube using several of the most common conditions related to duration since handwashing (unwashed, 15minutes post-wash, and 1-hour post-wash). Five replicates were undertaken per method, with the aim of developing a method for determining the best shedder test when limited data is available. Hierarchical Bayesian modelling (HBM) was used to compare five different modelling structures for the data. The most supported model was chosen and used to provide insights about the performance of each shedder test method and ultimately choose the best method for determining shedder status. The Bayesian modelling approach offers advantages in certain situations by providing posterior parameter distributions that avoid binary significance interpretations. The results from the models described in this study showed that Method 3 (1-hour hand wash) was best for shedder classification of those tested. This method most closely resembles personal natural behaviour while still allowing for some standardisation of the protocol. We use the data from this study to generate a shedder distribution and show how the knowledge of where an individual falls on the shedder distribution can be incorporated into an evaluation framework such as a Bayesian network that might be used in an evaluation of observations given activity level propositions.