Feasible algorithm for linear mixed model for massive data

作者：Zhao, Yanyan^*

来源：Communications in Statistics - Simulation and Computation, 2018, 47(4): 1126-1133.

DOI：10.1080/03610918.2017.1307395

• Linear mixed model
• Massive data
• Sequential singular value decomposition

This article studies computation problem in the context of estimating parameters of linear mixed model for massive data. Our algorithms combine the factored spectrally transformed linear mixed model method with a sequential singular value decomposition calculation algorithm. This combination solves the operation limitation of the method and also makes this algorithm feasible to big dataset, especially when the data has a tall and thin design matrix. Our simulation studies show that our algorithms make the calculation of linear mixed model feasible for massive data on ordinary desktop and have same estimating accuracy with the method based on the whole data.

• 出版日期2018
• 单位南开大学