A Perturbation Subsampling for Large Scale Data
Yujing Yao, Jin Zhezhen · Statistica Sinica · 2022
Rapid scientific and technological advances yield more and more large scale data.For the analysis of large scale data, subsampling methods and divide-and-conquer procedures are appealing as they ease computational burden while preserving the validity of the inference.Sampling with or without replacement is commonly used.In this paper, we propose a perturbation subsampling approach based on independent and identically distributed stochastic weights for the analysis of large scale data.We justify the method based on optimizing convex objective functions by establishing asymptotic consistency and normality for the resulting estimators.The method can provide consistent point estimator and variance estimator simultaneously.We present finite sample performance of the proposed method through simulation studies and real data analysis.