Multi-species Occupancy Models with Correlated Species Occurrences

This repository contains R code for

1. simulating data based on the assumptions of a multivariate probit occupancy model with correlated species occurrences

2. fitting the multi-species occupancy model and its latent-factor approximation to the simulated data

3. fitting a multivariate probit regression model to the matrix of simulated species occurrences

4. fitting the two multi-species occupancy models to the Tibetan Plateau data

The following publication contains a complete description of these models and the MCMC algorithms used to fit them.

Dorazio, Li, Xiao, and Kachel (2025) An evaluation of multi-species occupancy models with correlated species occurrences. Methods in Ecology and Evolution 16: 2681-2693.

The original MCMC algorithm used to fit the multivariate probit (MP) occupancy model (Appendix S1), included updates of Z_i,j that failed to properly account for the correlations in occurrence between species. The corrected MCMC algorithm is described in

Dorazio, Li, Xiao, and Kachel (2025) Correction to `An evaluation of multi-species occupancy models with correlated species occurrences'. Methods in Ecology and Evolution 17: in press.

AppendixS1_correctedMCMC.pdf

List of files

• MultispeciesOccModel_SimulateData.R - simulates data based on the assumptions of a multivariate probit (MP) occupancy model

• MultispeciesOccModel_MCMC.R - fits a multivariate probit (MP) occupancy model to data using an efficient MCMC algorithm

• MultispeciesOccModel_correctedMCMC.R - fits a multivariate probit (MP) occupancy model to data using the corrected MCMC algorithm

• MultispeciesOccModel_LV_JAGS.R - fits a latent factor (LF) occupancy model to data using JAGS

• MvProbitModel_MCMC.R - fits a multivariate probit regression model (MvP) to latent occurrences using an efficient MCMC algorithm

• SimulationStudy-Driver.R - calls functions to simulate a data set and then fit each model to this data set

• realdata.rds - RDS file containing the camera-trap survey data of mammalian species on the Tibetan Plateau

• FitMPmodelToTibetanData.R - driver for fitting the MP model to the Tibetan Plateau data

• FitLVmodelToTibetanData.R - driver for fitting the LF model to the Tibetan Plateau data

• EstimatePosteriorStats.R -- estimates posterior summary statistics from Markov chains

• AppendixS1_MCMC.pdf -- description of MCMC algorithm used to fit the multivariate probit (MP) occupancy model

• AppendixS1_correctedMCMC.pdf -- description of corrected MCMC algorithm used to fit the multivariate probit (MP) occupancy model

Required R packages

mvtnorm, tmvtnorm, msm, MASS, clusterGeneration, mcmcse, coda, rjags

Correspondence

Robert Dorazio (rmdorazio@gmail.com) and Lingyun Xiao (Lingyun.Xiao@xjtlu.edu.cn)