annmatrix (version 0.1.2)

pca: Principal Component Analysis for annmatrix Objects

Description

Performs principal component analysis on annmatrix objects and preserves row and column annotations in scores and loadings.

Usage

# S3 method for annmatrix
prcomp(
  x,
  retx = TRUE,
  center = TRUE,
  scale. = FALSE,
  tol = NULL,
  rank. = NULL,
  ...
)

Value

prcom object with rotation and x matrices turned into annmatrix

Arguments

x

annmatrix object.

retx

logical indicator whether the rotated variables should be returned (defaults to TRUE).

center

logical value indicating whether variables should be centered (defaults to TRUE).

scale.

logical value indicating whether variables should be scaled (defaults to FALSE).

tol

tolerance value indicating magnitude below which components will be omitted.

rank.

number specifying the maximal rank (max number of principal components to be used).

...

arguments passed to or from other methods.

Author

Karolis Koncevičius

Details

The resulting loadings ('rotation') and scores ('x') are turned into annmatrix objects with additional row and column annotations. Row annotations are copied from the original input matrix while column annotations contain computed information about each principal component. The added information contains: 1) principal component number, 2) standard deviation, 3) variance, 4) fraction of variance explained.

Examples

# construct annmatrix object
x <- matrix(rnorm(20*10), 20, 10)

coldata <- data.frame(group  = rep(c("case", "control"), each = 5),
                      gender = sample(c("M", "F"), 10, replace = TRUE))

rowdata <- data.frame(chr = sample(c("chr1", "chr2"), 20, replace = TRUE),
                      pos = runif(20, 0, 1000000))

X <- annmatrix(x, rowdata, coldata)

pca <- prcomp(t(X))
pca$rotation
pca$rotation$''

scores <- t(pca$rotation) %*% X
scores@var_explained
scores$gender