# riem.test2wass: Two-Sample Test with Wasserstein Metric In Riemann: Learning with Data on Riemannian Manifolds

 riem.test2wass R Documentation

## Two-Sample Test with Wasserstein Metric

### Description

Given M observations X_1, X_2, …, X_M \in \mathcal{M} and N observations Y_1, Y_2, …, Y_N \in \mathcal{M}, permutation test based on the Wasserstein metric (see riem.wasserstein for more details) is applied to test whether two distributions are same or not, i.e.,

H_0~:~\mathcal{P}_X = \mathcal{P}_Y

with Wasserstein metric \mathcal{W}_p being the measure of discrepancy between two samples.

### Usage

riem.test2wass(
riemobj1,
riemobj2,
p = 2,
geometry = c("intrinsic", "extrinsic"),
...
)


### Arguments

 riemobj1 a S3 "riemdata" class for M manifold-valued data. riemobj2 a S3 "riemdata" class for N manifold-valued data. p an exponent for Wasserstein distance \mathcal{W}_p (default: 2). geometry (case-insensitive) name of geometry; either geodesic ("intrinsic") or embedded ("extrinsic") geometry. ... extra parameters including npermthe number of permutations (default: 999). use.smootha logical; TRUE to use a smoothed Wasserstein distance, FALSE otherwise.

### Value

a (list) object of S3 class htest containing:

statistic

a test statistic.

p.value

p-value under H_0.

alternative

alternative hypothesis.

method

name of the test.

data.name

name(s) of provided sample data.

### Examples

#-------------------------------------------------------------------
#          Example on Sphere : a dataset with two types
#
# class 1 : 20 perturbed data points near (1,0,0) on S^2 in R^3
# class 2 : 30 perturbed data points near (0,1,0) on S^2 in R^3
#-------------------------------------------------------------------
## GENERATE DATA
mydata1 = list()
mydata2 = list()
for (i in 1:20){
tgt = c(1, stats::rnorm(2, sd=0.1))
mydata1[[i]] = tgt/sqrt(sum(tgt^2))
}
for (i in 1:20){
tgt = c(rnorm(1,sd=0.1),1,rnorm(1,sd=0.1))
mydata2[[i]] = tgt/sqrt(sum(tgt^2))
}
myriem1 = wrap.sphere(mydata1)
myriem2 = wrap.sphere(mydata2)

## PERFORM PERMUTATION TEST
#  it is expected to return a very small number, but
#  small number of 'nperm' may not give a reasonable p-value.

riem.test2wass(myriem1, myriem2, nperm=99, use.smooth=FALSE)

## Not run:
## CHECK WITH EMPIRICAL TYPE-1 ERROR
set.seed(777)
ntest = 1000
pvals = rep(0,ntest)

for (i in 1:ntest){
X = cbind(matrix(rnorm(30*2, sd=0.1),ncol=2), rep(1,30))
Y = cbind(matrix(rnorm(30*2, sd=0.1),ncol=2), rep(1,30))
Xnorm = X/sqrt(rowSums(X^2))
Ynorm = Y/sqrt(rowSums(Y^2))

Xriem = wrap.sphere(Xnorm)
Yriem = wrap.sphere(Ynorm)
pvals[i] = riem.test2wass(Xriem, Yriem, nperm=999)\$p.value
print(paste0("iteration ",i,"/",ntest," complete.."))
}

emperr = round(sum((pvals <= 0.05))/ntest, 5)
print(paste0("* EMPIRICAL TYPE-1 ERROR=", emperr))

## End(Not run)



Riemann documentation built on March 18, 2022, 7:55 p.m.