Composite-link generalized additive models for nested areal change of support:
coarse Poisson (or quasi-Poisson) counts linked to a fine intensity through a
composition matrix C, with anisotropic spatial P-splines and multi-resolution
covariate smooths (Cases A–C). Estimation is PIRLS + SOP. An optional
identifying restriction (orth.smooth = TRUE) defines residual spatial
heterogeneity after fine covariates.
The installable package lives in clgam/ (version 0.1.25).
No proprietary health data are shipped; examples use simulate_ata().
remotes::install_github("idaejin/CL-GAM", subdir = "clgam")From a local clone:
install.packages("clgam", repos = NULL, type = "source")
# or from inside clgam/:
install.packages(".", repos = NULL, type = "source")API, formula interface (s(x1, x2, ndx, bdeg, pord)), and examples:
clgam/README.md.
clgam::simulate_ata() (and simulate_ata_scenarios()) generate the nested
Voronoi DGPs used in the paper. Competitor Monte Carlo scripts for the full
simulation battery are not bundled in this repository; the package API is
sufficient to refit CL-GAM on the same designs.
Method name in papers: CL-GAM. R package: clgam.
- Ayma, Durban, Lee, Eilers (2016), Spatial Statistics (spatial CLMM, ATP/ATA)
- CL-GAM with multi-resolution covariates: manuscript in preparation