R/rNEF.r

In jinyung/otolith: Identification of fish otolith using digital images

#' Normalized elliptical Fourier analysis routine
#' 
#' @description A wrapper functions using the codes from J Claude's book,  
#'  for routine use to run normalized elliptical Fourier analysis.
#' @param A p x k x n array of raw semilandmarks configurations
#' @param plotpca logical. Whether to plot the PCA for quick preliminary view. 
#'  Plot first 3 PCs
#' @param class factor giving the species/grouping, used for plotting purpose, 
#'   e.g. using \code{sp} value from \code{\link{routine1}} or using the 
#'   \code{\link{getclass}} function. Used only when \code{plotpca=TRUE}
#' @param ... other arguments passed to \code{\link{plotpca}}
#' @return 
#'  \item{coeff}{matrix of all harmonic coefficients}
#'  \item{expvar}{explained variation of the NEF harmonics 
#'    (first harmonic is removed from the calculation, see reference)}  
#'  \item{size}{size of the semi-major axis of the first fitting ellipse. See
#'  reference}
#' @seealso 
#'  \code{\link{rGPA}}
#' @export
#' @references Claude J. (2008). \emph{Morphometrics with R}. Springer.

rNEF <- function (A, plotpca = FALSE, class, ...) {
  n <- dim(A)[3]
  p <- dim(A)[1]
  coeff <- matrix(NA, n, (p/2) * 4, 
                  dimnames = list(dimnames(A)[[3]], 
                                  c(paste0("A", 1:(p/2)), paste0("B", 1:(p/2)), 
                                    paste0("C", 1:(p/2)), paste0("D", 1:(p/2)))))
  size <- numeric()
  for (i in 1:n) {
    ef <- NEF(A[, , i])
    coeff[i, ] <- c(ef$A, ef$B, ef$C, ef$D)
    size[i] <- ef$size 
  }
  vharm <- apply(coeff, 2, var)
  variation <- apply(matrix(vharm, (p/2), 4), 1, sum)
  expvar <- round(cumsum(variation[-1]) / sum(variation[-1]), 3)[1:(p/2 - 1)]
  if (plotpca) {
    pca <- prcomp(coeff)
    plotpca(pca, size = size, class = class, ...)
  }
  invisible(list(coeff = coeff, expvar = expvar, size = size))
}