dML: Distance-based Machine Learning

In hk1785/MiVT: Microbiome Virtual Twins

Distance-based Machine Learning

Description

This function implements dML to predicts treatment effects.

Usage

dML(y, Tr, X, tree, n.folds = 10, n.rep = 2, alpha = seq(0.05, 0.95, 0.05), n.trees = 1000, n.neus = c(1/2, 1/3, 1/4))

Arguments

y	A vector of binary responses.
Tr	A vector of binary treatment status.
X	A feature (OTU or ASV) table where rows are features and columns are subjects.
tree	A rooted phylogenetic tree.
n.folds	The number of folds in the k-fold cross-validation (Default: 10).
n.rep	The number of repeats of the k-fold cross-validation (Default: 2).
alpha	Candidate alpha values that modulates the amount of L1 and L2 penalties to their linear combination for the elastic net (Default: seq(0.05, 0.95, 0.05)).
n.trees	The number of bagged decision trees for the random forest (Default: 1000).
n.neus	Candidate numbers of neurons on the first hidden layer for the three layer deep feedforward network: [M/2], [M/4], [M/8], [M/3], [M/6], [M/12], and [M/4], [M/8], [M/16], where M is the number of predictors (coordinates).

Value

$out.en$cv.cro: CV cross-entropy values for the elastic net and each distance measure. $out.en$Z: Predicted treatment effects using the elastic net.

$out.rf$cv.cro: CV cross-entropy values for the random forest and each distance measure. $out.rf$Z: Predicted treatment effects using the random forest.

$out.dfn$cv.cro: CV cross-entropy values for the deep feedforward network and each distance measure. $out.dfn$Z: Predicted treatment effects using the deep feedforward network.

$Z: Predicted treatment effects using dML.

Author(s)

Hyunwook Koh

References

Koh, H. Subgroup identification using virtual twins for human microbiome studies. (Under review).

Foster, J. C., Taylor, J. M. & Ruberg, S. J. Subgroup identification from randomized clinical trial data. Stat. Med. 30(24), 2867-2880 (2011).

Examples

data(fit)
data(tree)
data(tax.tab)

prop <- fit$pi
disp <- fit$theta

sim.biom <- gen.syn.dat(tree = tree, tax.tab = tax.tab, prop = prop, disp = disp)
sim.biom

qc.out <- biom.qc(biom = sim.biom)

dml.out <- dML(y = qc.out$sam.dat$y, Tr = qc.out$sam.dat$Tr, X = qc.out$otu.tab, tree = qc.out$tree)

hk1785/MiVT documentation built on Dec. 21, 2024, 1:08 p.m.