modelSelection: Find optimal common and distinctive components

Description Usage Arguments Value Author(s) See Also Examples

Description

Estimate the optimal number of common and distinctive components according to given selection criteria.

Usage

1
modelSelection(Input,Rmax,fac.sel,varthreshold=NULL,nvar=NULL,PCnum=NULL,center=FALSE,scale=FALSE,weight=FALSE, plot_common=FALSE, plot_dist=FALSE)

Arguments

Input

List of ExpressionSet objects, one for each block of data

Rmax

Maximum common components

fac.sel

PCA criteria for selection ("%accum", "single%", "rel.abs", "fixed.num")

varthreshold

Cumulative variance criteria for PCA selection. Threshold for "%accum" or "single%" criteria.

nvar

Relative variance criteria. Threshold for "rel.abs".

PCnum

Fixed number of components for "fixed.num".

center

Character (or FALSE) specifying which (if any) centering will be applied before analysis. Choices are "PERBLOCKS" (each block separately) or "ALLBLOCKS" (all data together).

scale

Character (or FALSE) specifying which (if any) scaling will be applied before analysis. Choices are "PERBLOCKS" (each block separately) or "ALLBLOCKS" (all data together).

weight

Logical indicating whether weighting is to be done. Choices are "BETWEEN-BLOCKS"

plot_common

Logical indicating whether to plot the explained variances (SSQ) of each block and its estimation and the ratios

plot_dist

Logical indicating whether to plot the explained variances (SSQ) and the accumulated variance for each block

Value

List containing:

common

List with common components results

commonComps

Optimal number of common components

ssqs

Matrix of SSQ for each block and estimator

pssq

ggplot object showing SSQ for each block and estimator

pratios

ggplot object showing SSQ ratios between each block and estimator

dist

List containg the results of distinct PCA for each input block; for each block PCAres and numComps is returned within a list

PCAres

List containing results of PCA, with fields "eigen", "var.exp", "scores" and "loadings"

nomComps

Number of components selected

Author(s)

Patricia Sebastian-Leon

See Also

omicsCompAnalysis

Examples

1
2
3
4
5
data(STATegRa_S3)
B1 <- createOmicsExpressionSet(Data=Block1.PCA,pData=ed.PCA,pDataDescr=c("classname"))
B2 <- createOmicsExpressionSet(Data=Block2.PCA,pData=ed.PCA,pDataDescr=c("classname"))
ms <- modelSelection(Input=list(B1, B2), Rmax=3, fac.sel="single\%", varthreshold=0.03, center=TRUE, scale=FALSE, weight=TRUE, plot_common=FALSE, plot_dist=FALSE)
ms

STATegRa documentation built on Nov. 8, 2020, 5:26 p.m.