RASCO

RASCO mitigates spatial confounding in cancer disease mapping to improve inference on covariate effects and spatial patterns of areal count data.

Key Features:

• Spatial Confounding Mitigation: Implements the RHZ, HH, and SPOCK approaches to alleviate correlation between spatial random effects and fixed covariates.
• Restricted Spatial Regressions: Constructs restricted spatial regressions by projecting latent spatial effects onto the orthogonal space of covariates or by displacing spatial locations.
• Model Flexibility: Supports parametric areal count models including Poisson, generalized Poisson, and negative binomial and quantifies spatial association using the conditional autoregressive (CAR) model.
• Bayesian Inference with INLA: Performs Bayesian inference and can be accelerated using the integrated nested Laplace approximation (INLA).
• Implementation: Implemented in R for fitting spatial and restricted spatial regression models for areal count responses.

Scientific Applications:

• Cancer disease mapping: Enables more accurate estimation of geographical patterns in cancer incidence, prevalence, status, and progression by addressing spatial confounding.
• Inference on covariate effects: Provides unbiased assessment of fixed covariate effects in areal count models when latent spatial association is present.

Methodology:

Implements RHZ, HH, and SPOCK approaches; constructs restricted spatial regressions via projection onto covariate-orthogonal space or spatial location displacement; fits Poisson, generalized Poisson, and negative binomial areal count models with CAR spatial association; performs Bayesian inference, often using INLA.