An Approach to Incorporate Subsampling into a Generic Bayesian Hierarchical Model

The goal of this paper is to provide a way for Bayesian statisticians to incorporate subsampling directly into the Bayesian hierarchical model of their choosing without imposing additional restrictive model assumptions. We are motivated by the fact that the rise of "big data" has created difficulties for statisticians to directly apply their methods to big datasets. We introduce a "data subset model" to the popular "data model, process model, and parameter model" framework used to summarize Bayesian hierarchical models. The hyperparameters of the data subset model are specified constructively in that they are chosen such that the implied size of the subset satisfies pre-defined computational constraints. Thus, these hyperparameters effectively calibrate the statistical model to the computer itself to obtain predictions/estimations in a pre-specified amount of time. Several properties of the data subset model are provided including: propriety, partial sufficiency, and semi-parametric properties. Simulated datasets will be used to assess the consequences of subsampling, and results will be presented across different computers to show the effect of the computer on the statistical analysis. Additionally, we provide a joint analysis of a high-dimensional dataset (roughly 10 gigabytes) consisting of 2018 5-year period estimates from the US Census Bureau's Public Use Micro-Sample (PUMS).