A0159
Title: Subsampling and rare events data beyond binary responses
Authors: HaiYing Wang - University of Connecticut (United States) [presenting]
Abstract: Rare event data occur when certain events occur with very small probabilities. Subsampling effectively reduces the computational cost of analyzing rare event data without losing significant estimation efficiency. Existing investigations on subsampling with rare event data focus on binary response models. Rare event data beyond binary responses are investigated. There will be no statistical efficiency loss if sufficient data points for the non-rare observations are sampled. In the scenario of estimation efficiency loss due to downsampling, an optimal sampling design is developed to minimize the information loss.
