Two Sample Test with Distribution-Valued Data

Share:

Description

Two Sample Test with Distribution-Valued Data

Usage

1
2
d3m(cases, control, rm.mean = FALSE, rm.var = FALSE, paranum = 101,
  q = 2, bsn = 5000, seed = 100)

Arguments

cases

name of case group data (matrix)

control

names of control group data (matrix)

rm.mean

standarize each rows of cases and control to mean=0.

rm.var

standarize each rows of cases and control to var=1.

paranum

the number of quatile discretization + 1. Default is discretized by 1 %.

q

power of Wasserstein metric. Default is q = 2.

bsn

the number of resampling. Default is bsn = 5000.

seed

seed for random number generator.

Details

this function is designed for two sample test based on Wasserstein metric. The function computes the the p-values based Wasserstein metric and resampling method. If rm.mean=F and rm.var=F, then statistical test is performed only based on more than 3rd order moments.

Value

pval p-value.

test.stat test statistic.

cases case group data used in the statistical test.

control control group data used in the statistical test.

Author(s)

Yusuke Matsui & Teppei Shimamura

References

Yusuke Matsui, Masahiro Mizuta, Satoru Miyano and Teppei Shimamura.(2015) D3M:Detection of differential distributions of methylation patterns (submitted). BIORXIV/2015/023879.

Antonio Irpino and Rossanna Verde.(2015) Basic Statistics for distributional symbolic variables: a new metric-based approach. Adv.Data.Anal.Classif(9) 143–175

Examples

1
2
3
4
nrep <-12
cases <- Map(rbeta,rep(30,nrep),rep(1,nrep),rep(5,nrep)); cases <- do.call("rbind",cases)
control <- Map(rbeta,rep(30,nrep),rep(1,nrep),rep(5,nrep)); control <- do.call("rbind",control)
d3m(cases,control,paranum = 101, q = 2, bsn = 1000,seed = 100)