The Poisson lognormal model and variants can be used for a variety of multivariate problems when count data are at play (including PCA, LDA and network inference for count data). This package implements efficient algorithms to fit such models accompanied with a set of functions for vizualisation and diagnostic.

## Installation

PLNmodels is available on CRAN. The development version is available on Github.

### R Package installation

#### CRAN dependencies

PLNmodels needs the following CRAN R packages, so check that they are are installed on your computer.

required_CRAN <- c("R6", "glassoFast", "Matrix", "Rcpp", "RcppArmadillo",
"nloptr", "igraph", "grid", "gridExtra", "dplyr",
"tidyr", "ggplot2", "corrplot", "magrittr", "devtools")
not_installed_CRAN <- setdiff(required_CRAN, rownames(installed.packages()))
if (length(not_installed_CRAN) > 0) install.packages(not_installed_CRAN)

#### Bioconductor dependencies

PLNmodels also needs two BioConductor packages

required_BioC <- c("phyloseq", "biomformat")
not_installed_BioC <- setdiff(required_BioC, rownames(installed.packages()))
if (length(not_installed_BioC) > 0) BiocManager::install(not_installed_BioC)

#### Installing PLNmodels

• For the last stable version, use the CRAN version
install.packages("PLNmodels")
• For the development version, use the github install
remotes::install_github("jchiquet/PLNmodels")
• For a specific tagged release, use
remotes::install_github("jchiquet/PLNmodels@tag_number")

## Usage and main fitting functions

The package comes with an ecological data set to present the functionality

library(PLNmodels)
data(trichoptera)
trichoptera <- prepare_data(trichoptera$Abundance, trichoptera$Covariate)

The main fitting functions work with the usual R formula notations, with mutivariate responses on the left hand side. You probably want to start by one of them. Check the corresponding vignette and documentation page. There is a dedicated vignettes for each model in the package (See http://jchiquet.github.io/PLNmodels/articles/).

### Unpenalized Poisson lognormal model (aka PLN)

myPLN <- PLN(Abundance ~ 1, data = trichoptera)

### Rank Contraint Poisson lognormal for Poisson Principal Component Analysis (aka PLNPCA)

myPCA <- PLNPCA(Abundance ~ 1, data = trichoptera, ranks = 1:8)

### Poisson lognormal discriminant analysis (aka PLNLDA)

myLDA <- PLNLDA(Abundance ~ 1, grouping = Group, data = trichoptera)

### Sparse Poisson lognormal model for sparse covariance inference for counts (aka PLNnetwork)

myPLNnetwork <- PLNnetwork(Abundance ~ 1, data = trichoptera)

## References

Please cite our work using the following references:

• J. Chiquet, M. Mariadassou and S. Robin: Variational inference for probabilistic Poisson PCA, the Annals of Applied Statistics, 12: 2674–2698, 2018. link

• J. Chiquet, M. Mariadassou and S. Robin: Variational inference for sparse network reconstruction from count data, Proceedings of the 36th International Conference on Machine Learning (ICML), 2019. link