funPCA: Functional Principal Component Analysis (FPCA) via...

In statKim/robfpca: Functional principal component analysis for partially observed elliptical process

Functional Principal Component Analysis (FPCA) via Conditional Expectation

Description

FPCA is performed via PACE (Principal Analysis via Conditional Expectation) proposed by Yao et al. (2005).

Usage

funPCA(Lt, Ly, mu, cov, sig2 = 0, work.grid = NULL, K = NULL, PVE = 0.99)

Arguments

Lt	a list of vectors or a vector containing time points for all curves
Ly	a list of vectors or a vector containing observations for all curves
mu	a mean function estimated at work.grid
cov	a covariance function estimated at work.grid
sig2	a noise variance
work.grid	a work grid
K	a number of FPCs. If K is NULL, K is selected by PVE.
PVE	a proportion of variance explained

Value

a list contatining as follows:

data	a list containing Lt and Ly
lambda	the first K eigenvalues
eig.fun	the first K eigenvectors
pc.score	the first K FPC scores
K	a number of FPCs
PVE	a proportion of variance explained
work.grid	a work grid
eig.obj	an object of the eigenanalysis
mu	a mean function
cov	a covariance function
sig2	a noise variance

Examples

### Generate example data
set.seed(100)
x.list <- sim_delaigle(n = 100,
                       type = "partial",
                       out.prop = 0.2,
                       dist = "normal")
x <- list2matrix(x.list)

### Estimate robust covariance function
work.grid <- seq(0, 1, length.out = 51)
cov.obj <- cov_ogk(x,
                   type = "huber",
                   bw = 0.1)

### Functional principal component analysis
fpca.obj <- funPCA(Lt = x.list$Lt,
                   Ly = x.list$Ly,
                   mu = cov.obj$mean,
                   cov = cov.obj$cov,
                   work.grid = work.grid,
                   PVE = 0.95)
fpc.score <- fpca.obj$pc.score

# ### Give same result in the above
# fpca.obj <- robfpca.partial(x,
#                             type = "huber",
#                             PVE = 0.95,
#                             bw = 0.1)
# fpc.score <- fpca.obj$pc.score

statKim/robfpca documentation built on April 15, 2023, 10:12 p.m.