wdm::Indep_test Class Reference

#include <wdm.hpp>

Public Member Functions
	Indep_test (std::vector< double > x, std::vector< double > y, std::string method, std::vector< double > weights=std::vector< double >(), bool remove_missing=true, std::string alternative="two-sided", std::vector< int > seeds=std::vector< int >(), bool y_continuous=true)
std::string	method () const
	Returns the requested method name.
std::string	alternative () const
	Returns the requested alternative hypothesis.
double	n_eff () const
	Returns Kish's effective sample size after missing-value removal.
double	estimate () const
	Returns the estimated dependence measure.
double	statistic () const
	Returns the method-specific transformed test statistic.
double	p_value () const
	Returns the asymptotic p-value.

Detailed Description

Asymptotic independence test based on a dependence measure.

The test stores the estimate, transformed test statistic, effective sample size, and p-value. Weighted transformations use Kish's effective sample size. The approximation must have enough effective observations for the selected method.

Available methods:

• "pearson", "prho", "cor": Pearson correlation
• "spearman", "srho", "rho": Spearman's \( \rho \)
• "kendall", "ktau", "tau": Kendall's \( \tau \)
• "blomqvist", "bbeta", "beta": Blomqvist's \( \beta \)
• "hoeffding", "hoeffd", "d": Hoeffding's \( D \)
• "chatterjee", "cxi", "xi": Chatterjee's \( \xi \)

Hoeffding's D supports only the two-sided alternative. Other methods support "two-sided", "less", and "greater".

Note

Weighted analytic inference for Chatterjee's xi assumes that the weights are fixed or depend only on x, the normalized weights are diffuse, and y is continuous. The weighted estimate remains available when y is tied, but analytic inference with unequal weights does not. For a continuous response, estimate() reports the general denominator-corrected coefficient while statistic() standardizes the analytically covered approximation \(1 - 3 A\).

Constructor & Destructor Documentation

Indep_test()

wdm::Indep_test::Indep_test ( std::vector< double >  x, std::vector< double >  y, std::string  method, std::vector< double >  weights = std::vector<double>(), bool  remove_missing = true, std::string  alternative = "two-sided", std::vector< int >  seeds = std::vector<int>(), bool  y_continuous = true  )

inline

Constructs and evaluates an independence test.

Parameters

x,y	input vectors of equal length.
method	the dependence measure; see class details for possible values.
weights	optional finite, nonnegative case weights with positive total weight.
remove_missing	if true, rows containing a NaN are removed; otherwise a NaN raises an exception.
alternative	indicates the alternative hypothesis and must be one of "two-sided", "greater" or "less"; "greater" corresponds to positive association, "less" to negative association. For Hoeffding's \( D \), only "two-sided" is allowed. The natural one-sided alternative for Chatterjee's xi is "greater".
seeds	optional seeds for random Chatterjee predictor-tie breaking.
y_continuous	whether the Chatterjee response distribution is known to be continuous. Set this to false for a discrete response even if the sample has no observed response ties. Observed ties always override this value.

Exceptions

std::runtime_error

for invalid inputs, method or alternative names, unsupported Hoeffding alternatives, or unavailable weighted Chatterjee inference for a discrete or tied response.

---

The documentation for this class was generated from the following file:

• include/wdm.hpp