pcalg (version 2.2-4)

gmInt: Graphical Model 8-Dimensional Interventional Gaussian Example Data

Description

This data set contains a matrix with an ensemble of observational and interventional data from eight Gaussian variables. The corresponding (data generating) DAG model is also stored.

Usage

data(gmInt)

Format

The format is a list of four components

x:

Matrix with 5000 rows (one row a measurement) and 8 columns (corresponding to the 8 variables

targets:

List of (mutually exclusive) intervention targets. In this example, the three entries integer(0), 3 and 5 indicate that the data set consists of observational data, interventional data originating from an intervention at vertex 3, and interventional data originating from an intervention at vertex 5.

target.index:

Vector with 5000 elements. Each entry maps a row of x to the corresponding intervention target. Example: gmInt$target.index[3322] == 2 means that x[3322, ] was simulated from an intervention at gmInt$targets[[2]], i.e. at vertex 3.

g:

Formal class 'graphNEL' [package "graph"] with 6 slots, representing the true DAG from which observational and interventional data was sampled.

Source

The data set is identical to the one generated by

  set.seed(40)
  p <- 8
  n <- 5000
  gGtrue <- randomDAG(p, prob = 0.3)
  pardag <- as(gGtrue, "GaussParDAG")
  pardag$set.err.var(rep(1, p))
  targets <- list(integer(0), 3, 5)
  target.index <- c(rep(1, 0.6*n), rep(2, n/5), rep(3, n/5)) x1 <- rmvnorm.ivent(0.6*n, pardag)
  x2 <- rmvnorm.ivent(n/5, pardag, targets[[2]], 
                      matrix(rnorm(n/5, mean = 4, sd = 0.02), ncol = 1))
  x3 <- rmvnorm.ivent(n/5, pardag, targets[[3]], 
                      matrix(rnorm(n/5, mean = 4, sd = 0.02), ncol = 1))
  gmInt <- list(x = rbind(x1, x2, x3), 
                targets = targets, 
                target.index = target.index, 
                g = gGtrue)
  

Details

The data was generated as indicated below. First, a random DAG model was generated, then 5000 samples were drawn from this model: 3000 observational ones, and 1000 each from an intervention at vertex 3 and 5, respectively (see gmInt$target.index).

Examples

data(gmInt)
str(gmInt, max = 3)
pairs(gmInt$x, gap = 0, pch = ".")