iCARH.model: Runs the integrative CAR Horseshoe model

In iCARH: Integrative Conditional Autoregressive Horseshoe Model

Description

Infers treatment effects, association with heterogeneous omic variables, pathway perturbation among other parameters (e.g. time dependence). Regression coefficients (beta parameter) are initialized using a univariate regression ignoring time and metabolite dependence.

Usage

iCARH.model(
  X,
  Y = NULL,
  drug,
  groups = NULL,
  pathways,
  tau = 1.2,
  NA_value = -99999,
  init = T,
  ...
)

Arguments

X	the metabolomics time-course data with dimensions timepoints x observations x variables
Y	the additional omic time-course data with dimensions timepoints x observations x variables
drug	treatment effect. Could be either continuous (an administered drug or other external factor) or binary (cases vs controls). In the binary case the groups argument can be safely removed. NA values not allowed in drug. Dimensions are timepoints x observations
groups	grouping vector (binary). Use when drug is continuous.
pathways	pathway adjacency matrices as returned by iCARH.getPathwaysMat
tau	global sparsity parameter τ as in Jendoubi, T., & Ebbels, T. (2018)
NA_value	NA values are incompatible with stan. This is a wrapper to encode missing values in X and Y. NAs will be replaced by NA_value and will be inferred (only for X and Y data).
init	If TRUE use iCARH provided initialization function. Passed to Stan otherwise. Please see Stan manual on init possible values.
...	additional stan parameters

Value

stan object

Examples

data.sim = iCARH.simulate(4, 8, 10, 2, 2, path.probs=0.3, Zgroupeff=c(0,4),
beta.val=c(1,-1,0.5, -0.5))
XX = data.sim$XX
Y = data.sim$Y
Z = data.sim$Z
pathways = data.sim$pathways

rstan_options(auto_write = TRUE)
options(mc.cores = 2)
fit = iCARH.model(XX, Y, Z,groups=rep(c(0,1), each=4), pathways,
 control = list(adapt_delta = 0.99, max_treedepth=10), iter = 2, chains = 2)

iCARH documentation built on Aug. 28, 2020, 1:10 a.m.