Precise asymptotics for linear mixed models is a topic that the speaker became involved with, against the grain of his other research, because of his December 2018 visit to Stanford University. It has since led to papers in Biometrika and Journal of the Royal Statistics Society, Series B. A separate query from Stanford University in 2022 led to new work on crossed, rather than nested, random effects. In this talk the speaker will present some recent results from his 2024 sabbatical on the crossed case. In addition to the usual selling points of approximate inference, optimal design and sample size calculations, a new pay-off will be discussed: estimator behavior. In particular, intriguing similarities between maximum likelihood estimators for linear mixed models with nested and crossed random effects will be exposed.
This research has involved collaborations and discussions with Aishwarya Bhaskaran, Apratim Dey, Swarnadip Ghosh, Jiming Jiang, Iain Johnstone, Luca Maestrini, Song Mei and Art Owen.