mmp3 / zinLDA

R package for Zero-Inflated Latent Dirichlet Allocation

zinLDA: Zero-Inflated Latent Dirichlet Allocation

About

Zero-inflated latent Dirichlet allocation is an unsupervised, hierarchical, generative probabilistic model that facilitates dimensionality reduction and detection of sparse latent clusters. This package provides implementation of a Markov chain Monte Carlo (MCMC) sampling procedure for the zinLDA model. Additionally, it provides a method for simulating sparse count data from an underlying zinLDA model.

While the original paper developed this model for applications to microbiome data and microbial subcommunity detection, it is flexible enough to be used with numerous types of discrete count data.

Installation

You can install the latest version of zinLDA from GitHub with:

install.packages("devtools")
devtools::install_github("rebeccadeek/zinLDA")

Documentation and Examples

Help documentation for the zinLDA package is available in R. After installing the package from GitHub via devtools and loading it with library() use ? to access the documentation for any of the four main functions in the package. E.g.

?zinLDA

Additionally, the zinLDA package contains a vignette with a more detailed description of the model, how it differs from currently existing methods, and examples on how to simulate data from zinLDA and fit the model. To include the vignette during installation and to access it please use:

devtools::install_github("rebeccadeek/zinLDA", build_vignettes = TRUE)
vignette(package = "zinLDA")

Alternatively, there is a companion website that contains the same statistical details and examples as the vignette.

Contact

To report any bugs, issues, or suggestions please use the issue feature on GitHub or contact the maintainer Rebecca Deek via email.