Translating Big Data into Smart Data for Veterinary Epidemiology

VanderWaal, Kimberly; Morrison, Robert B.; Neuhauser, Claudia; Vilalta, Carles; Perez, Andres M.

and connectivity (8) , and limited availability of trained personnel capable of extracting data from databases (14) . Finally, high velocity data create a need to automate data pipelines for routine and repeated use. Automation is key for harnessing big data for monitoring and surveillance (8, 49, 56) . A major criticism of big data analytics is that it lacks the rigor of hypothesis-driven, controlled experiments for determining causation (16, 62) . However, correlations identified through the analysis of big data are useful for hypothesis generation and prediction (5, 17) . In addition, the increasing numbers of measurable explanatory factors available from diverse sources necessitate the use of relatively new (to veterinary epidemiologists) statistical approaches, such as machine learning, that are more appropriate for handing datasets with a large number of covariates (17, 18, (63) (64) (65) . Due to the large number of potential variables, care must be taken to identify spurious correlations (5, 66) , and the use of large datasets does not necessarily increase a study's validity. Sound epidemiological principles for the interpretation of observational data are required (62) .