Reassessing the Paradigms of Statistical Model-Building

Abstract

Statistical model-building is the science of constructing models from data and from information about the data-generation process, with the aim of analysing those data and drawing inference from that analysis. Many statistical tasks are undertaken during this analysis; they include classification, forecasting, prediction and testing. Model-building has assumed substantial importance, as new technologies enable data on highly complex phenomena to be gathered in very large quantities. This creates a demand for more complex models, and requires the model-building process itself to be adaptive. The word “paradigm” refers to philosophies, frameworks and methodologies for developing and interpreting statistical models, in the context of data, and applying them for inference. In order to solve contemporary statistical problems it is often necessary to combine techniques from previously separate paradigms. The workshop addressed model-building paradigms that are at the frontiers of modern statistical research. It tried to create synergies, by delineating the connections and collisions among different paradigms. It also endeavoured to shape the future evolution of paradigms.