Variational Multivariate Spatial Small Area Estimation R package vmsae version 0.1.2 Variational Autoencoded Multivariate Spatial Fay-Herriot models are designed to efficiently estimate population parameters in small area estimation. This package implements the variational generalized multivariate spatial Fay-Herriot model (VGMSFH) using 'NumPyro' and 'PyTorch' backends, as demonstrated by Wang, Parker, and Holan (2025) <doi:10.48550/arXiv.2503.14710>. The 'vmsae' package provides utility functions to load weights of the pretrained variational autoencoders (VAEs) as well as tools to train custom VAEs tailored to users specific applications. Software Depends: R (>= 3.5.0) Imports: dplyr, ggplot2, gridExtra, sf, tidyr, reticulate, methods, rlang Zhenhua Wang <zhenhua.wang@missouri.edu> Comprehensive R Archive Network (CRAN) Zhenhua Wang [aut, cre], Paul A. Parker [aut, res], Scott H. Holan [aut, res] MIT + file LICENSE (https://CRAN.R-project.org/package=vmsae/LICENSE) 2025-10-08 application/tgz https://CRAN.R-project.org/package=vmsae doi:10.32614/CRAN.package.vmsae