A Framework for Mediation Analysis with Massive Data

During recent years, mediation analysis has become increasingly popular in many research fields. Basically, the aim of mediation analysis is to investigate the direct effect of exposure on outcome together with indirect effects along the pathways from exposure to outcome. There has been a great number of articles that applied mediation analysis to data from hundreds or thousands of individuals. With the rapid development of technology, the volume of avaliable data increases exponentially, which brings new challenges to researchers. It is often computationally infeasible to directly conduct statistical analysis for large datasets. However, there are very few results on mediation analysis with massive data. In this paper, we propose to use the subsampled double bootstrap as well as divide-and-conquer algorithm to perform statistical mediation analysis for large-scale dataset. Extensive numerical simulations are conducted to evaluate the performance of our method. Two real data examples are also provided to illustrate the usefulness of our approach in practical application.