simGibbs: Generate binary (*Y*, *A*, *C*) from chain graph model under...

In netchain: Inferring Causal Effects on Collective Outcomes under Interference

Description

Generate binary (Y, A, C) from chain graph model under simplest scenario.

Usage

simGibbs(
  n.unit,
  n.gibbs,
  n.sample,
  weight.matrix,
  treat.matrix,
  cov.matrix,
  init.prob = 0.5,
  treat.prob = 0.5,
  cov.prob = 0.5,
  n.burn = 100,
  yvalues = c(1, 0)
)

Arguments

n.unit	the number of units (m).
n.gibbs	the number of independent Gibbs sampler.
n.sample	the number of samples from each Gibbs sampling (n = n.gibbs x n.sample).
weight.matrix	a [m x m] symmetric relational matrix where W_ij = 1 indicates the existence of undirected edges between Y_i and Y_j and its magnitude. Here W_ii represents the main effect of Y_i.
treat.matrix	a [m x m] matrix where treat.matrix_ij indicates the magnitude of direct effect from A_i to Y_j.
cov.matrix	a [m x m] matrix where treat.matrix_ij indicates the magnitude of direct effect from C_i to Y_j.
init.prob	an initial probability generating (Y, A, C) from Bernoulli distribution.
treat.prob	a probability updating A in Gibbs sampling procedure.
cov.prob	a probability updating C in Gibbs sampling procedure.
n.burn	the number of burn-in sample in Gibbs sampling (≥ n.obs).
yvalues	a vector of distinct binary outcome values. Defaults to c(0,1).

Value

a list of [n.gibbs] x [n.sample] independent observations:

inputY	a [([n.gibbs] x [n.sample]) x m] matrix for outcomes.
inputA	a [([n.gibbs] x [n.sample]) x m] matrix for treatments.
inputC	a [([n.gibbs] x [n.sample]) x m] matrix for confounders.

Examples

library(netchain)
weight.matrix <- matrix(c(0.5, 1, 0, 1, 0.3, 0.5, 0, 0.5, -0.5), 3, 3)
simobs <- simGibbs(n.unit = 3, n.gibbs = 200, n.sample = 10, 
                  weight.matrix,
                  treat.matrix = 0.5*diag(3), cov.matrix= (-0.3)*diag(3) )
inputY <- simobs$inputY
inputA <- simobs$inputA
inputC <- simobs$inputC

netchain documentation built on Feb. 17, 2020, 9:11 a.m.