nonpar: Nonparametric two-sample testing using kernel-based test...
In neurodata/graphstats: Graph Statistics

Description Usage Arguments Value Author(s) References

This is a simple implementation of the kernel-based test statistic for the nonparametric two-sample testing problem of given X_1, X_2, …, X_n i.i.d. F and Y_1, Y_2, …, Y_m i.i.d. G, test the null hypothesis of F = G against the alternative hypothesis of F \not = G. The test statistic is based on embedding F and G into a reproducing kernel Hilbert space and then compute a distance between the resulting embeddings. For this primitive, the Hilbert space is associated with the Gaussian kernel.

1 2	nonpar(G1, G2, dim = NULL, sigma = NULL, alpha = 0.05, bootstrap_sample = 200, verbose = FALSE)

`dim`	dimension of the latent position that graphs are embeded into. Defaults to `NULL` which selects the optimal dimensionality using .
`sigma`	bandwidth of the rbf kernel for computing test statistic
`alpha`	Significance level of hypothesis testing. Defaults to `.05`.
`bootstrap_sample`	Number of bootstrap samples when performing hypothesis tesing
`verbose`	logical indicating whether to print output to console. Defaults to `FALSE`.
`g1`	an igraph object
`g2`	an igraph object

T A scalar value T such that T is near 0 if the rows of X and Y are from the same distribution and T far from 0 if the rows of X and Y are from different distribution.

Eric Bridgeford ericwb95@gmail.com, Youngser Park youngser@jhu.edu, Kemeng Zhang kzhang@jhu.edu.

Tang, M., Athreya, A., Sussman, D.L., Lyzinski, V., Priebe, C.E. A nonparametric two-sample hypothesis testing problem for random graphs

neurodata/graphstats documentation built on May 14, 2019, 5:19 p.m.