Skip to content

Naive, constrained resampling

If you need to truncate the furnishing distributions of your uncertain datasets before applying a causality test, use the following method.

# CausalityToolsBase.causality — Method.

causality(source::AbstractUncertainIndexValueDataset, 
    target::AbstractUncertainIndexValueDataset, 
    test::ConstrainedTest{CT, CR}) where {CT, CR}

Apply a causality test of type CT to test.n independent realisations of source and target, after first constraining the supports of the uncertain values furnishing the datasets.

See also ConstrainedTest.

source

ConstrainedTest

# CausalityTools.CausalityTests.ConstrainedTest — Type.

ConstrainedTest(test::CausalityTest, constraints::ConstrainedResampling, n::Int)
ConstrainedTest(test::CausalityTest, constraints::ConstrainedResampling)

A causality test where the supports of the uncertain values furnishing the uncertain values in the datasets are truncated (according to the provided constraints) prior to performing the test.

n controls the number of independent resamplings over which the test is performed (if not provided, n is set to 1 by default).

Examples

Assume we want to apply a causality test to two uncertain datasets X and Y, but restricting the ranges of values their elements can take while resampling. 

# Constraints for X.indices and X.values
cx = (TruncateQuantiles(0.3, 0.7), TruncateStd(1.5))

# Constraints for Y.indices and Y.values
cy = (TruncateQuantiles(0.3, 0.7), TruncateStd(1.5))

# Need to gather to feed to the ConstrainedTest constructor.
cs = ConstrainedIndexValueResampling(cx, cy)

# A cross mapping test applied to 100 independent realisations
# of `X` and `Y`, resampled after constraining the data.
ctest = ConstrainedTest(CrossMappingTest(), cs, 100)

source