mdine: Microbiome differential network estimation

In kevinmcgregor/mdine: Microbiome Differential Network Estimation

Description

Microbiome differential network estimation

Usage

mdine(
  Y,
  X,
  Z,
  lambda = NULL,
  offset = NULL,
  mc.cores = chains,
  iter = 1000,
  chains = 4,
  quant = c(0.025, 0.975),
  nnet.MaxNWts = NULL,
  ...
)

Arguments

Y	The (unrarefied) taxa count matrix with rows as samples and columns as taxa. The last column is the reference category, and is not included in the estimated network.
X	The model matrix (including an intercept column)
Z	Vector containing the binary variable over which the network is assumed to vary.
lambda	Network penalization parameter. If NULL, then lambda is estimated
offset	A vector containing an offset term for each subject
mc.cores	The number of cores to run MCMC chains in parallel
iter	The number of MCMC iterations. By default the first half of the iterations will be used as warmup.
chains	The number of MCMC chains.
quant	Vector containing the lower and upper quantiles of the posterior distribution to create credible intervals.
nnet.MaxNWts	Numeric specifying the maximum number of weights in the nnet::multinom call. This might need to be increased for a large number of OTUs. Defaults to 1000.
...	Other arguments passed to rstan::sampling

Details

MDiNE is a model based on multinomial logistic regression to estimate precision matrix-based networks within two groups.

Value

An object of class mdine containing posterior means for the model parameters, credible intervals, and the stanfit object.

stan.fit	The object returned from rstan. This can be used for MCMC diagnostics provided in the rstan package
post_mean	List contatining estimated posterior means for the model parameters
ci	List contatining credible intervals for all parameters
lam_mle	Initial value of lambda used as mean in the prior distribution for lambda

References

McGregor, Labbe, and Greenwood 2019: DOI

Examples

library(mdine)
data(crohns)

X <- model.matrix(~disease, data=crohns$covars)
md.fit <- mdine(Y=crohns$otu.counts, X=X, Z=X[,2], mc.cores=1, iter=1000)

kevinmcgregor/mdine documentation built on April 10, 2021, 4:18 a.m.