gapStatistics: gap statistics

In Caleb-Huo/IS-Kmeans: Integrative Sparse Kmeans

Description

gap statistics to tune gamma

Usage

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

Arguments

d	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.
gamma	Penalty on total number of features. Larger gamma will yeild small number of selected features.
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.

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

set.seed(123)

# Generate two random omics datasets
mu <- c(-3,1,3)
Simu1_mRNA <- rbind(cbind(matrix(rnorm(40*5, mu[1], 0.1),40,5),
                          matrix(rnorm(40*5, mu[2], 0.1),40,5),
                          matrix(rnorm(40*5, mu[3], 0.1),40,5)),
                    matrix(rnorm(10*15,0,0.1),10,15))

mu <- c(1,3,-3)
Simu1_methyl <- rbind(cbind(matrix(rnorm(40*5, mu[1], 0.1),40,5),
                            matrix(rnorm(40*5, mu[2], 0.1),40,5),
                            matrix(rnorm(40*5, mu[3], 0.1),40,5)),
                      matrix(rnorm(10*15,0,0.1),10,15))

## feature modules across two datasets
group <- list(c(1:10,51:60), c(11:20,61:70), c(21:30,71:80), c(31:40,81:90))

DList <- rbind(Simu1_mRNA, Simu1_methyl)
dim(DList)

d <- t(DList)

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

Caleb-Huo/IS-Kmeans documentation built on May 6, 2019, 9:27 a.m.