| shrinkt.stat {st} | R Documentation |
shrinkt.stat and shrinkt.fun compute the ``shrinkage t'' statistic
of Opgen-Rhein and Strimmer (2007).
shrinkt.stat(X, L, var.equal=TRUE, verbose=TRUE) shrinkt.fun(L, var.equal=TRUE, verbose=TRUE)
X |
data matrix. Note that the columns correspond to variables (``genes'') and the rows to samples. |
L |
group indicator vector. Samples belonging to the first group are assigned a `1', and those belonging to the second group a `2'. |
var.equal |
assume equal (default) or unequal variances in each group. |
verbose |
print out some (more or less useful) information during computation. |
The ``shrinkage t'' statistic is similar to the usual t statistic, with the replacement of the sample variances by corresponding shrinkage estimates. These are derived in a distribution-free fashion and with little a priori assumptions. Despite its simplicity, it can be shown that ``shrinkage t'' procduces highly accurate rankings - see Opgen-Rhein and Strimmer (2007).
shrinkt.stat returns a vector containing the ``shrinkage t''
statistic for each variable/gene.
The corresponding shrinkt.fun functions return a function that
produces the ``shrinkage t'' statistics when applied to a data matrix
(this is very useful for simulations).
Rainer Opgen-Rhein (http://opgen-rhein.de) and Korbinian Strimmer (http://strimmerlab.org).
Opgen-Rhein, R., and K. Strimmer. 2007. Accurate ranking of differentially expressed genes by a distribution-free shrinkage approach. Statist. Appl. Genet. Mol. Biol. 6:9. (http://www.bepress.com/sagmb/vol6/iss1/art9/)
studentt.stat,
diffmean.stat,
efront.stat,
sam.stat,
samL1.stat,
modt.stat.
# load st library
library("st")
# load Choe et al. (2005) data
data(choedata)
X <- choe2.mat
dim(X) # 6 11475
L <- choe2.L
L
# shrinkage t statistic (equal variances)
score = shrinkt.stat(X, L)
order(abs(score), decreasing=TRUE)[1:10]
# [1] 10979 11068 50 1022 724 5762 43 4790 10936 9939
# lambda.var (variance vector): 0.3882
# shrinkage t statistic (unequal variances)
score = shrinkt.stat(X, L, var.equal=FALSE)
order(abs(score), decreasing=TRUE)[1:10]
# [1] 11068 50 10979 724 43 1022 5762 10936 9939 9769
# lambda.var (variance vector): 0.3673 0.3362
# compute q-values and local false discovery rates
# note the procedure automatically estimates the degree of freedom
# so any statistic following a t distribution may be supplied
library("fdrtool")
fdr.out = fdrtool(score, statistic="studentt")
sum( fdr.out$qval < 0.05 )
sum( fdr.out$lfdr < 0.2 )
fdr.out$param