Robertson, E.P., Walsh, D.P., Martin, J., Work, T.M., Kellogg, C.A., Evans, J.S., Barker, V., Hawthorn, A., Aeby, G., Paul, V.J., Walker, B.K., Kiryu, Y., Woodley, C.M., Meyer, J.L., Rosales, S.M., Studivan, M., Moore, J.F., Brandt, M.E., and Bruckner, A., 2023, Expert assessments of hypotheses concerning the etiological agent(s) of Stony Coral Tissue Loss Disease collected during a rapid prototyping project: U.S. Geological Survey data release, https://doi.org/10.5066/P9DLNEBY.
We collected this data from experts to elicit, structure, and evaluate existing knowledge regarding the etiology of stony coral tissue loss disease.This dataset was used in a Bayesian hierarchical modelling framework to formalize the process for combining expert knowledge regarding disease etiology that (a) identifies competing hypotheses about the agent(s) causing disease; (b) quantifies the belief weights for each hypothesis using two expert elicitation techniques; and (c) presents techniques to update the belief weight. We note that these data results may be sensitive to the expertise of the participants, their initial background level of knowledge, their unconscious decision biases (e.g., pet hypotheses), and results also may be influenced by the number of experts and the time they invest into the elicitation (e.g., the time they invest into reading each publication). There may also be some linguistic uncertainty among experts (e.g., understanding the hypotheses). For M1 we only included 15 experts and for M2 we only used five studies. We intend for the M2 results to only be used as a simple demonstration of these methods rather than for interpretation of SCTLD etiology due to the limited number of studies. Ideally, many more studies will need to be incorporated before these results can be used for conservation decision making. These studies could also be gathered over longer time periods (e.g., years to decades) and belief weights updated as new papers are published on the topic.Similarly, more experts or more rounds of elicitation may benefit inference from M1.
