Compositionally-warped additive mixed modeling for a wide variety of non-Gaussian spatial data

As with the advancement of geographical information systems, non-Gaussian spatial data sets are getting larger and more diverse. This study develops a general framework for fast flexible regression, especially spatial/spatiotemporal modeling. The developed model, termed compositionally-warped additive mixed model (CAMM), combines an (AMM) Gaussian process to wide variety continuous including other effects. A specific advantage proposed CAMM is that it requires no explicit assumption distribution unlike existing AMMs. Monte Carlo experiments show estimation accuracy computational efficiency modeling fat-tailed and/or skewed distributions. Finally, applied crime examine empirical performance regression analysis prediction. result shows provides intuitively reasonable coefficient estimates outperforms AMM in terms prediction accuracy. verified be potentially covers