LDAcov

LDAcov implements a modified Latent Dirichlet Allocation (LDA) model in R that integrates covariates to identify latent communities in multivariate biodiversity data and infer environmental drivers of community change.

Key Features:

• Incorporation of Covariates: Integrates environmental and external covariates directly into the LDA framework to link latent communities to drivers of change.
• Inference and Prediction: Supports spatial interpolation of community composition and prediction under future environmental-change scenarios.
• Model Estimation via MCMC: Employs Markov Chain Monte Carlo (MCMC) algorithms to estimate model parameters, including the number of groups, from multivariate count data.

Scientific Applications:

• Ecological change analysis: Used to investigate effects of global change phenomena such as climate change and habitat fragmentation on biodiversity.
• High-dimensional biodiversity data: Applied to multivariate datasets with many species across locations and time points to identify latent community structure.
• Southeastern Amazonian forest studies: Applied in experimental studies of repeated fires and forest fragmentation to reveal impacts on plant assemblages.

Methodology:

Uses a modified LDA model that integrates covariates and applies MCMC algorithms to estimate model parameters and the number of groups, as demonstrated on simulated datasets.