exVarPlot: Visualize explained variance of PC-1 and PC-2 of a matrix...

In LionelRohner/LRTools: A collection of functions written for analyzing Illumina HM450 and EPIC data

Description

To be honest, I am not entirely sure how informative this function is. I used it to see how much the explained variance of the PCs varies in the lower tail in order to choose a suitable number of the most variable CpG-probes. This number can be quite arbitrary, and if this number is in a region of the distribution that fluctuates a lot, it may be wiser to reconsider the number of most variable rows. Explained variance is calculated as the ratio of eigenvalues of PC-1 or PC-2 divided by the sum of all eigenvalues as proposed here: https://stats.stackexchange.com/questions/254592/calculating-pca-variance-explained/254598.

Usage

exVarPlot(mat, lowerLim, upperLim, stepSize = 10, onlyPC1 = F)

Arguments

mat	A matrix of numeric values, where rows are probes, genes, metabolites, etc. and the columns are samples.
lowerLim	The lower limit of the most variable row entries.
upperLim	The upper limit of the most variable row entries.
stepSize	The step size. By default the step size is 10.
onlyPC1	Only consider PC-1. Is "FALSE" by default.

Value

A plot and a data.frame with the explained variance of PC-1 or PC-1 + PC-2 at each step.

Examples

A <- data.frame(rnorm(1000,0.2,0.01),runif(1000,0.2,0.75),rnorm(1000,0.75,0.05),
                          rnorm(1000,0.2,0.01),runif(1000,0.2,0.6),rnorm(1000,0.6,0.05))

rownames(A) <- paste("cg",sample(10000000:99999999,1000,replace = FALSE), sep ="")
colnames(A) <- rep(paste("Sample_",1:6,sep=""))

A <- as.matrix(A)

exVarPlot(mat=as.matrix(A),
          lowerLim = 10,
          upperLim = 1000)

LionelRohner/LRTools documentation built on Dec. 17, 2021, 1:10 a.m.