predict.gpca: Predict generalized PCA scores or reconstruction on new data

In andland/generalizedPCA: Generalized PCA

Description

Predict generalized PCA scores or reconstruction on new data

Usage

## S3 method for class 'gpca'
predict(object, newdata, type = c("PCs", "link", "response"),
  ...)

Arguments

object	generalized PCA object
newdata	matrix of the same exponential family as in object. If missing, will use the data that object was fit on
type	the type of fitting required. type = "PCs" gives the PC scores, type = "link" gives matrix on the natural parameter scale and type = "response" gives matrix on the response scale
...	Additional arguments

Examples

# construct a low rank matrices in the natural parameter space
rows = 100
cols = 10
set.seed(1)
loadings = rnorm(cols)
mat_np = outer(rnorm(rows), rnorm(cols))
mat_np_new = outer(rnorm(rows), loadings)

# generate a count matrices
mat = matrix(rpois(rows * cols, c(exp(mat_np))), rows, cols)
mat_new = matrix(rpois(rows * cols, c(exp(mat_np_new))), rows, cols)

# run Poisson PCA on it
gpca = generalizedPCA(mat, k = 1, M = 4, family = "poisson")

PCs = predict(gpca, mat_new)

andland/generalizedPCA documentation built on May 12, 2019, 2:42 a.m.