Moment-assisted Subsampling based Maximum Likelihood Estimator
share
This paper proposes a moment-assisted subsampling method which can improve the estimation efficiency of existing subsampling estimators. The motivation behind this approach stems from the fact that sample moments can be efficiently computed even if the sample size of the whole data set is huge. Through the generalized method of moments, this method incorporates informative sample moments of the whole data into the subsampling estimator. The moment-assisted estimator is asymptotically normal and has a smaller asymptotic variance compared to the corresponding estimator without incorporating sample moments of the whole data. The asymptotic variance of the moment-assisted estimator depends on the specific sample moments incorporated. Under the uniform subsampling probability, we derive the optimal moment that minimizes the resulting asymptotic variance in terms of Loewner order. Moreover, the moment-assisted subsampling estimator can be rapidly computed through one-step linear approximation. Simulation studies and a real data analysis were conducted to compare the proposed method with existing subsampling methods. Numerical results show that the moment-assisted subsampling method performs competitively across different settings. This suggests that incorporating the sample moments of the whole data can enhance existing subsampling technique.
READ FULL TEXT