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.
You can install the latest version of zinLDA
from GitHub with:
install.packages("devtools")
devtools::install_github("rebeccadeek/zinLDA")
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.
To report any bugs, issues, or suggestions please use the issue feature on GitHub or contact the maintainer Rebecca Deek via email.