DNNMP: Discrete nearest-neighbor mixture process

This repository contains the R package dnnmp (currently developer's version) and R scripts to reproduce the numerical results in

Xiaotian Zheng. Athanasios Kottas. Bruno Sansó. Zheng, X., Kottas, A., & Sansó, B. (2023). Bayesian geostatistical modeling for discrete-valued processes. Environmetrics, 34(7), e2805. https://doi.org/10.1002/env.2805

Installing and using the dnnmp package

You can install the package with devtools

devtools::install_github("xzheng42/dnnmp-examples-env-2023", subdir = "dnnmp")
library(dnnmp)

Main functions of the package are dnnmp and predict.dnnmp:

• dnnmp fits a DNNMP model via Markov chain Monte Carlo (MCMC).
• predict.dnnmp (or simply predict) generates posterior predictive samples for a set of new locations given a dnnmp object.

Notes: The current version was tested on macOS 10.15.7 under R version 4.2.2 and on Fedora Linux 38 under R version 4.3.2.

Workflow to reproduce numerical results

R scripts to reproduce results in the paper are available in data-examples/, and data/ contains the survey routes (names, numbers, location information, status, etc.) available in the North American Breeding Bird Survey (BBS) Dataset and the northern cardinal (Cardinalis cardinalis) count data.

• Run all simulation experiments: run_all_sim_rscripts.R.

• Run all BBS data examples: run_all_bbs_rscripts.R.

• Simulation experiments in Section 5.1 and Section D.1: sim1_mcmc.R and sim1_results (first simulation experiment); sim2_mcmc.R, sim2_model_comp, sim2_results (second simulation experiment);

• Data analyses in Section 5.3: bbs_data_analysis.R.

• Data analyses in Section D.2: bbs_gaus_sa_1_mcmc.R, bbs_gaus_sa_2_mcmc.R, bbs_gaus_sa_1_results.R, and bbs_gaus_sa_2_results.R (Section D.2.1); bbs_comp_1_mcmc.R and bbs_comp_1_results.R (Section D.2.2); bbs_comp_2_mcmc.R and bbs_comp_2_results.R (Section D.2.3);

• Randomized quantile residual analysis for model checking: sim_model_checking.R and bbs_model_checking.R.