R/fitLoessKD.matrix.R

In aroma.cn: Copy-Number Analysis of Large Microarray Data Sets

setMethodS3("fitLoessKD", "matrix", function(X, Y, ...) {
  # Argument 'X':
  X <- as.matrix(X)

  # Argument 'Y':
  Y <- as.matrix(Y)
  if (!all(dim(Y) == dim(X))) {
    dimXStr <- paste(dim(X), collapse="x")
    dimYStr <- paste(dim(Y), collapse="x")
    throw("The dimensions of argument 'Y' and 'X' do not match: ",
          dimYStr, " != ", dimXStr)
  }

  # Record input data
  data <- list(X=X, Y=Y)

  # Drop missing values
  ok <- rowAlls(is.finite(X) & is.finite(Y))
  X <- X[ok,,drop=FALSE]
  Y <- Y[ok,,drop=FALSE]

  # Estimate relationship for each dimension
  fitList <- list()
  for (cc in seq_len(ncol(X))) {
    fitList[[cc]] <- loess(Y[,cc] ~ X)
  }

  # Create predict function
  predictY <- function(X, ...) {
    # Argument 'X':
    if (ncol(X) != length(fitList)) {
      throw("The number of columns in argument 'X' does not match the number of fitted dimensions: ", ncol(X), " != ", length(fitList))
    }

    X <- as.matrix(X)
    naValue <- NA_real_
    Y <- array(naValue, dim=dim(X))
    ok <- rowAlls(is.finite(X))
    ok <- which(ok)
    X <- X[ok,,drop=FALSE]
    for (cc in seq_len(ncol(X))) {
      fit <- fitList[[cc]]
      yPred <- predict(fit, newdata=X)
      Y[ok,cc] <- yPred
    }
    Y
  } # predictY()

  fit <- list(data=data, fitList=fitList, predictY=predictY)

  class(fit) <- c("LoessKDFit", class(fit))

  fit
})

 
setMethodS3("normalizeLoessKD", "matrix", function(X, fit, ...) {
  # Argument 'fit':
  fit <- Arguments$getInstanceOf(fit, "LoessKDFit")

  fit$predictY(X, ...)
})


setMethodS3("fitLoessKD", "data.frame", function(X, ...) {
  X <- as.matrix(X)
  fitLoessKD(X, ...)
})

setMethodS3("normalizeLoessKD", "data.frame", function(X, ...) {
  XN <- X
  X <- as.matrix(X)
  XN <- normalizeLoessKD(X, ...)
  XN <- as.data.frame(XN)
  XN
})


##############################################################################
# HISTORY
# 2010-09-19 [HB]
# o Created.
##############################################################################