bigpca (version 1.1)

pca.scree.plot: Make scree plots for any PCA

Description

Make a scree plot using eigenvalues from princomp(), prcomp(), svd(), irlba(), big.PCA(), etc. Note that most these return values which need to be squared to be proper eigenvalues. There is also an option to use the estimate.eig.vpcs() function to estimate any missing eigenvalues (e.g, if using a function like irlba' to calculate PCA) and then to visualise the fitline of the estimate on the scree plot.

Usage

pca.scree.plot(eigenv, elbow = NA, printvar = TRUE, min.dim = NA,
  M = NULL, add.fit.line = FALSE, n.xax = max(30, length(eigenv)),
  linear = TRUE, verbose = FALSE, return.data = FALSE, ...)

Arguments

eigenv

the vector of eigenvalues actually calculated

elbow

the number of components which you think explain the important chunk of the variance of the dataset, so further components are modelled as reflecting noise or very subtle effects, e.g, often the number of components used is decided by the 'elbow' in a scree plot (see 'pca.scree.plot')

printvar

logical, whether to print summary of variance calculations

min.dim

the size of the smaller dimension of the matrix submitted to singular value decomposition, e.g, number of samples - i.e, the max number of possible eigenvalues, alternatively use 'M'.

M

optional enter the original dataset 'M'; simply used to derive the dimensions, alternatively use 'min.dim'.

add.fit.line

logical, if there is an existing scree plot, adds the fit line from this estimate to the plot ('pca.scree.plot' can use this option using the parameter of the same name)

n.xax

number of components to include on the x-axis

linear

whether to use a linear model to model the 'noise' eigenvalues; alternative is a 1/x model with no intercept.

verbose

logical, whether to display additional output

return.data

logical, whether to return the percentages of variance explained for each component, or nothing (just plot)

...

further arguments to the plot function

Value

Either a vector of variance percentages explained, or nothing (just a plot), depending on value of 'return.data'

See Also

pca.scree.plot

Examples

# NOT RUN {
require(irlba)
nsamp <- 100; nvar <- 300; subset.size <- 25; elbow <- 6
mat <- matrix(rnorm(nsamp*nvar),ncol=nsamp) 
#this gives the full solution
pca <- svd(mat,nv=subset.size,nu=0)
pca2 <- irlba(mat,nv=subset.size,nu=0)
# show alternate fits for linear versus 1/x fit
pca.scree.plot((pca2$d^2)[1:subset.size],n.xax=100,add.fit.line=TRUE,
               min.dim=min(dim(mat)),linear=TRUE, elbow=6, ylim=c(0,1400))
pca.scree.plot((pca2$d^2)[1:subset.size],n.xax=100,add.fit.line=TRUE,
              min.dim=min(dim(mat)),linear=FALSE, elbow=40, ylim=c(0,1400))
subset.size <- 75
pca2 <- irlba(mat,nv=subset.size,nu=0)
pca.scree.plot((pca2$d^2)[1:subset.size],n.xax=100,add.fit.line=TRUE,
              min.dim=min(dim(mat)),linear=TRUE, elbow=6, ylim=c(0,1400))
pca.scree.plot((pca2$d^2)[1:subset.size],n.xax=100,add.fit.line=TRUE,
              min.dim=min(dim(mat)),linear=FALSE, elbow=40, ylim=c(0,1400))
# }