R/computeAbility.R

In integIRTy: Integrating Multiple Modalities of High Throughput Assays Using Item Response Theory

computeAbility <-
function(respMat, dscrmn=dscrmn, dffclt=dffclt, c=rep(0, length(dffclt)), parallel=FALSE)
{
  d=rep(1, length(dffclt))
  method='BM'
# item parameter input matrix
  itemParMat <- cbind(dscrmn, dffclt, c, d)
# extract unique pattern:
  PatternStr <- apply(respMat, 1, paste, collapse='')
  uniquePatternStr <- unique(PatternStr)
  unique2allInd <- match(uniquePatternStr, PatternStr)
  all2uniqueInd <- match(PatternStr, uniquePatternStr)
  uniquePatternMat <- matrix(as.numeric(unlist(strsplit(uniquePatternStr, ''))),
                             ncol=ncol(respMat), byrow=TRUE)
  thetaEstFunc <- function(i) {
    thetaEst(itemParMat, uniquePatternMat[i, ],
             method=method, range=c(-8,8))
  }
  i <- NULL # quiet R CMD check
  if(parallel==TRUE){
    abilityEst <- foreach(i=1:nrow(uniquePatternMat), .combine='c', .packages='integIRTy') %dopar% thetaEstFunc(i)
  } else {
    abilityEst <- foreach(i=1:nrow(uniquePatternMat), .combine='c') %do% thetaEstFunc(i)
  }
  names(abilityEst) <- uniquePatternStr
	
  allScore <- abilityEst[all2uniqueInd]
  names(allScore) <- rownames(respMat) # respMat contains no NA, hence rowname is corresponding gene name
  #list(uniqueScore=abilityEst, unique2allInd=unique2allInd, all2uniqueInd=all2uniqueInd, compTime=compTime, allScore=allScore)
  allScore
}