Big data and biostatistics: The death of the asymptotic Valhalla

Despite the ubiquity of Big Data in the modern scientific discourse, most references describe storage and query considerations and rarely full-flexed analyses. In this article, we propose another definition with particular relevance to biometrics. We argue that the complexity of the generating measure of biological process means that the model complexity of any statistical model will have to be smaller. Only, when the model is used for prediction can we have any hope that the number of available features reasonably outnumbers the desired complexity of the model.