gapStatistics: gap statistics

In Caleb-Huo/MIS-Kmeans:

Description

gap statistics to tune gamma

Usage

gapStatistics(d, K = 3, B = 10, gammas = NULL, alpha = 1,
  group = NULL, seed = 15213, silence = FALSE)

Arguments

d	A list of S studies, each study is a combined data matrix n*J, where n is number of subjects, J=J1+J2+... and J1 is number of features in omics dataset 1 and J2 is number of features in omics dataset 2...
K	number of clusters
B	number of permutations.
alpha	balance between group sparsity and individual sparsity. alpha=1 yeilds no group sparsity. alpha=0 yeilds no individual penalty.
group	Prior group information. Potentially these group can contain overlap features. group is a list and each element of the list is feature index.
seed	random seed
silence	do not print progress in silence = TRUE.
gamma	Penalty on total number of features. Default is given. Larger gamma will yeild small number of selected features.

Details

gap statistics to tune gamma, the tuning parameter to control number of features.

Value

a table. Each row represents a gamma.

gamma	input gammas
score	objective score: see obj0 in ISKmeans function
E.score	mean value of permutated score
se.score	standard error of permutated score
gapStat	gap statistics, score - E.score
numF	number of selected features

Author(s)

Caleb

Examples

S <- 2
K <- 3
G <- 1000
g1 <- 50
g2 <- 50
n0 <- 20
n <- K*n0
labels <- cut(1:n,breaks=K,labels=FALSE)

set.seed(32611)
S1 <- matrix(rnorm(G*n), nrow=G, ncol=n)
S2 <- matrix(rnorm(G*n), nrow=G, ncol=n)

S1[1:g1, labels==1] <- S1[1:g1, labels==1] + 2
S1[1:g1, labels==3] <- S1[1:g1, labels==3] - 2
S1[g1 + 1:g2, labels==1] <- S1[g1 + 1:g2, labels==1] - 2
S1[g1 + 1:g2, labels==2] <- S1[g1 + 1:g2, labels==2] + 2

S2[1:g1, labels==2] <- S2[1:g1, labels==2] + 2
S2[1:g1, labels==1] <- S2[1:g1, labels==1] - 2
S2[g1 + 1:g2, labels==2] <- S2[g1 + 1:g2, labels==2] - 2
S2[g1 + 1:g2, labels==3] <- S2[g1 + 1:g2, labels==3] + 2

S = list(t(S1),t(S2))
groups <- Map('c',1:g1,g1 + 1:g2)

gapResult <- gapStatistics(d=S,K=3,B=10,group=groups)
print(gapResult)

Caleb-Huo/MIS-Kmeans documentation built on May 17, 2019, 2:45 p.m.