R/fitWHLAPLM.matrix.R

In aroma.core: Core Methods and Classes Used by 'aroma.*' Packages Part of the Aroma Framework

# Fit the log-additive model assuming *homoscedastic* error terms.
# Each data element can be given a weight.  Moreover, if there
# are missing values, these will be given zero weights.
setMethodS3("fitWLAPLM", "matrix", function(y, ...) {
  # Explicit call to avoid method dispatching overheads.
  fitWLAPLM.matrix(y, ..., maxIterations=1)
})




setMethodS3("fitWHLAPLM", "matrix", function(y, ...) {
  K <- nrow(y)
  I <- ncol(y)
  y <- log2(y)
  thetaIdxs <- seq_len(I)
  phiIdxs <- I+seq_len(K)

  fit <- fitWHRCModel.matrix(y, ...)

  est <- fit$Estimates
  se <- fit$StdErrors

  # Chip effects
  thetaIdxs <- 1:I
  beta <- est[thetaIdxs]
  theta <- 2^beta

  # Probe affinities
  phiIdxs <- (I+1):(I+K)
  alpha <- est[phiIdxs]
  alpha[K] <- -sum(alpha[1:(K-1)])
  phi <- 2^alpha

  # The RMA model is already fitted with constraint prod(phi) = 1.
  # No rescaling needed.

  seTheta <- 2^(se[thetaIdxs])
  sePhi <- 2^(se[phiIdxs])

  fit$theta <- theta
  fit$seTheta <- seTheta
  fit$phi <- phi
  fit$sePhi <- sePhi

  fit
}) # fitWHLAPLM()