edit

Null Hypothesis Significance Testing

NHST is used to determine whether the observed value of a network measure is larger, or smaller, or similar to, what is expected by chance. The function to perform a test is typically applied after you have generated randomized networks using either null models or network permutations.

# EcologicalNetwork.test_network_propertyFunction.

Null Hypothesis Significance Testing

test_network_property(N::EcoNetwork, f, S; test::Symbol=:greater)

Test whether the observed value (through applying a function f) on an empirical network N differs from the distribution derived from measuring the same value on a collection of randomized networks S. S is an array of networks of the same type as N.

There are two possible values for the test keyword argument: :greater and :smaller. The test is one-tailed. The results are returned as a NetworkTestOutput object (see ?EcologicalNetwork.NetworkTestOutput for the complete edocumentation).

The p-value (pval) is measured by counting the proportion of networks with a larger (resp. smaller) value of the measure than the original network, as in normal permutation tests.

The original value of the measure is given (v0), as well as the z-scores (z) of all randomized networks.

source

The output is a NetworkTestOutput object, with a number of fields.

# EcologicalNetwork.NetworkTestOutputType.

Output of a permutation-based test

  • pval – the test p-value
  • test – the type of test (:smaller or :greater)
  • v0 – the measure of the empirical network
  • n – the number of randomized networks used
  • hits – the number of randomized network matching the test condition
  • z – the z-scores of the statistics for each randomized network

source