Year: 2020
Type: article
Abstract: We derive and present explicit algorithms to facilitate streamlined computing\nfor variational inference for models containing higher level random effects.\nExisting literature, such as Lee and Wand (2016), is such that streamlined\nvariational inference is restricted to mean field variational Bayes algorithms\nfor two-level random effects models. Here we provide the following extensions:\n(1) explicit Gaussian response mean field variational Bayes algorithms for\nthree-level models, (2) explicit algorithms for the alternative variational\nmessage passing approach in the case of two-level and three-level models, and\n(3) an explanation of how arbitrarily high levels of nesting can be handled\nbased on the recently published matrix algebraic results of the authors. A\npay-off from (2) is simple extension to non-Gaussian response models. In\nsummary, we remove barriers for streamlining variational inference algorithms\nbased on either the mean field variational Bayes approach or the variational\nmessage passing approach when higher level random effects are present.
Authors: Nolan, TH, Menictas, M, Wand, MP
Institution: University of Technology Sydney
Language: English
Subfield: Statistics and Probability
Field: Mathematics
Domain: Physical Sciences
Open Access status: green
OpenAlex ID: W4288418830