RFgroove (version 1.1)

fpca: Functional Principal Component Analysis

Description

Compute the functional PCA from a set of curves.

Usage

fpca( x, nbasisInit, propVar = 0.9, reconstruct = FALSE,  varName = NULL, verbose = FALSE)

Arguments

x

The set of curves.

nbasisInit

The number of initial spline coefficients.

propVar

The proportion of explained variance.

reconstruct

Should the data be reconstruct after dimension reduction ?

varName

The name of the current functional variable.

verbose

Should the details be printed.

Value

A list with two components:

design

The matrix of the principal components ;

smoothData

The reconstructed data if ‘reconstruct == TRUE’.

Details

The Functional PCA is performed in two steps. First we express each discretized curves as a linear combination of ‘nbasisInit’ spline basis functions. Then a multivariate PCA is computed on the spline coefficients. The final number of principal components is chosen such that the proportion of explained variance is larger than ‘propVar’.

References

Ramsay, J. O., and Silverman, B. W. (2006), Functional Data Analysis, 2nd ed., Springer, New York.

See Also

hardThresholding

Examples

  data(toyRegFD)
  x <- toyRegFD$FDlist[[1]]
  PCs <- fpca(x=x, nbasisInit=32, propVar=.9, reconstruct=TRUE)
  plot(x[1,])
  lines(PCs$smoothData[1,], col=2)