CondIndTest function

Wrapper function for conditional independence tests.

Tests the null hypothesis that Y and E are independent given X.

CondIndTest(Y, E, X, method = "KCI", alpha = 0.05,
  parsMethod = list(), verbose = FALSE)

Arguments

• Y: An n-dimensional vector or a matrix or dataframe with n rows and p columns.
• E: An n-dimensional vector or a matrix or dataframe with n rows and p columns.
• X: An n-dimensional vector or a matrix or dataframe with n rows and p columns.
• method: The conditional indepdence test to use, can be one of "KCI", "InvariantConditionalQuantilePrediction", "InvariantEnvironmentPrediction", "InvariantResidualDistributionTest", "InvariantTargetPrediction", "ResidualPredictionTest".
• alpha: Significance level. Defaults to 0.05.
• parsMethod: Named list to pass options to method.
• verbose: If TRUE, intermediate output is provided. Defaults to FALSE.

Returns

A list with the p-value of the test (pvalue) and possibly additional entries, depending on the output of the chosen conditional independence test in method.

Examples

# Example 1
set.seed(1)
n <- 100
Z <- rnorm(n)
X <- 4 + 2 * Z + rnorm(n)
Y <- 3 * X^2 + Z + rnorm(n)
test1 <- CondIndTest(X,Y,Z, method = "KCI")
cat("These data come from a distribution, for which X and Y are NOT
cond. ind. given Z.")
cat(paste("The p-value of the test is: ", test1$pvalue))

# Example 2
set.seed(1)
Z <- rnorm(n)
X <- 4 + 2 * Z + rnorm(n)
Y <- 3 + Z + rnorm(n)
test2 <- CondIndTest(X,Y,Z, method = "KCI")
cat("The data come from a distribution, for which X and Y are cond.
ind. given Z.")
cat(paste("The p-value of the test is: ", test2$pvalue))

References

Please cite C. Heinze-Deml, J. Peters and N. Meinshausen: "Invariant Causal Prediction for Nonlinear Models", arXiv:1706.08576

and the corresponding reference for the conditional independence test.