Matrix-Variate Non-Gaussian Linear Regression Models R package MVNGmod version 0.1.2 Fits matrix-variate variance-gamma (MVVG) and matrix-variate normal-inverse-Gaussian (MVNIG) linear regression models using expectation-conditional maximization (ECM) algorithms. The models accommodate clustered matrix-valued responses, with unequal numbers of observations across subjects, correlated responses, skewness, and within-subject dependence. Functions are provided for model fitting, prediction, and subject-level influence assessment using approximate generalized Cook's distances. The package also includes motivating periodontal data from Gullah-speaking African Americans with Type-II diabetes. For details on the underlying matrix-variate distributions (MVVG and MVNIG), see Gallaugher and McNicholas (2019, <doi:10.1016/j.spl.2018.08.012>). Software Depends: R (>= 3.5.0) Imports: Bessel, clusterGeneration, DistributionUtils, matlib, maxLik, truncnorm, pracma, matrixcalc, purrr Suggests: knitr, rmarkdown Samuel Soon <samksoon2@gmail.com> Comprehensive R Archive Network (CRAN) Samuel Soon [aut, cre], Dipankar Bandyopadhyay [aut], Qingyang Liu [aut] MIT + file LICENSE (https://CRAN.R-project.org/package=MVNGmod/LICENSE) 2026-07-16 application/tgz https://CRAN.R-project.org/package=MVNGmod doi:10.32614/CRAN.package.MVNGmod