Knowledge-based bias correction : a case study in veterinary decision support

In collaboration with experts from veterinary research institutes throughout Europe, we developed a decision-support system for the early detection of Classical Swine Fever in pigs. For evaluating our system's diagnostic performance, practitioners and researchers collected data from the real-world field and from laboratory experiments. Originating from different sources, these data could not be viewed as constituting an unbiased sample from a single probability distribution. In this paper, we present a knowledge-based method for correcting the biases in estimates from such divergent data. We demonstrate the use of our method for estimating the sensitivity and specificity characteristics of our veterinary decision-support system.

ISBN:9781614994183

Publication year:2014

Accessibility:Open