eofRecon: EOF reconstruction (Empirical Orthogonal Functions analysis)

In marchtaylor/sinkr: Collection of functions with emphasis in multivariate data analysis

EOF reconstruction (Empirical Orthogonal Functions analysis)

Description

This function reconstructs the original field from an EOF object of the function eof.

Usage

eofRecon(EOF, pcs = NULL, newpcs = NULL, uncenter = TRUE, unscale = TRUE)

Arguments

EOF	An object resulting from the function eof.
pcs	The principal components (PCs) to use in the reconstruction (defaults to the full set of PCs: pcs=seq(ncol(EOF$u)))
newpcs	An (optional) matrix of new principal coordinates to use in the reconstruction. This would typically come from a gappy dataset whose missing values are to be predicted based on the EOF loadings of a the EOF object (see eofPred).
uncenter	Logical. Should reconstructed matrix be un-centered (e.g. if centered = TRUE was used in linkeof). Default is TRUE.
unscale	Logical. Should reconstructed matrix be un-centered (e.g. if scaled = TRUE was used in linkeof). Default is TRUE.

Examples

set.seed(1)
iris.gappy <- as.matrix(iris[,1:4])
iris.gappy[sample(length(iris.gappy), 0.25*length(iris.gappy))] <- NaN
Er <- eof(iris.gappy, method="svd", recursive=TRUE) # recursive (RSEOF)
Enr <- eof(iris.gappy, method="svd", recursive=FALSE) # non-recursive (LSEOF)
iris.gappy.recon.r <- eofRecon(Er)
iris.gappy.recon.nr <- eofRecon(Enr)

# Reconstructed values vs. observed values
op <- par(mfrow=c(1,2))
lim <- range(iris.gappy, na.rm=TRUE)
plot(iris.gappy, iris.gappy.recon.r, 
col=c(2:4)[iris$Species], main="recursive=TRUE", xlim=lim, ylim=lim)
abline(0, 1, col=1, lwd=2)
plot(iris.gappy, iris.gappy.recon.nr, 
col=c(2:4)[iris$Species], main="recursive=FALSE", xlim=lim, ylim=lim)
abline(0, 1, col=1, lwd=2)
par(op)

# Reconstructed values from gappy data vs. all original values
op <- par(mfrow=c(1,2))
plot(as.matrix(iris[,1:4]), iris.gappy.recon.r, 
col=c(2:4)[iris$Species], main="recursive=TRUE")
abline(0, 1, col=1, lwd=2) 
plot(as.matrix(iris[,1:4]), iris.gappy.recon.nr, 
col=c(2:4)[iris$Species], main="recursive=FALSE")
abline(0, 1, col=1, lwd=2) 

marchtaylor/sinkr documentation built on July 4, 2022, 5:48 p.m.