riem.test2wass: Two-Sample Test with Wasserstein Metric

Description Usage Arguments Value Examples

View source: R/inference_test2wass.R

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

1
2
3
4
5
6
7
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

nperm

the number of permutations (default: 999).

use.smooth

a 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

 1
 2
 3
 4
 5
 6
 7
 8
 9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
#-------------------------------------------------------------------
#          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 June 20, 2021, 5:07 p.m.