R/intIRTeasyRun.R

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

intIRTeasyRun <- function(platforms, model=3, guessing=FALSE,
                          addPermutedScore=FALSE, fold=1,
                          echo=TRUE, parallel=FALSE){
  nPlat <- length(platforms)
	########################################
	# fit on individual platform
	########################################
  if(echo==TRUE){
    cat('Performing ltm fit for each platform', '\n')	
  }	
	#fits <- list('vector')
  fitFunc <- function(i) fitOnSinglePlat(data=platforms[[i]],
                                         model=model, guessing=guessing)	
	## fits: a list of list
  plat <- NULL # to quite R CMD check
  if(parallel==TRUE){
    fits <- foreach(plat = 1:nPlat, .packages='integIRTy') %dopar% fitFunc(plat)
  } else {
    fits <- foreach(plat = 1:nPlat) %do% fitFunc(plat)
  }	
	########################################
	# calculate latent trait estimates from each platform
	########################################
  if(echo==TRUE){
    cat('Performing latent trait estimation for each platform', '\n')	
  }
	# assemble the item parameters and input matrix as a list: each individual platform followed by integrated data
  i <- NULL # to quite R CMD check
  a_list <- foreach(i=1:nPlat) %do% coef(fits[[i]]$fit)[, 'Dscrmn']; a_list[[length(a_list)+1]] <- c(unlist(a_list)) # append integrated parameter
  b_list <- foreach(i=1:nPlat) %do% coef(fits[[i]]$fit)[, 'Dffclt']; b_list[[length(b_list)+1]] <- c(unlist(b_list)) # append integrated parameter
  tempf <- function(i) {
    if(guessing==TRUE) {
      coef(fits[[i]]$fit)[, 'Gussng']
    } else {
      rep(0, ncol(platforms[[i]]))
    }
  }	
  i <- NULL # to quite R CMD check
  c_list <- foreach(i=1:nPlat) %do% tempf(i); c_list[[length(c_list)+1]] <- c(unlist(c_list)) # append integrated parameter
  platform_list <- foreach(i=1:nPlat) %do% platforms[[i]]; 
  platform_list[[length(platform_list)+1]] <- foreach(i=1:nPlat, .combine='cbind') %do% platforms[[i]]
  compScoreFunc <- function(i){
    computeAbility(platform_list[[i]], 
                   dscrmn=a_list[[i]],
                   dffclt=b_list[[i]], 
                   c=c_list[[i]])
  }
	# this is a matrix each column corresponding to a platform in the original order
  plat <- NULL # to quite R CMD check
  if(parallel==TRUE){
    estimatedScoreMat <- foreach(plat = 1:length(platform_list), .combine='cbind', .packages='integIRTy') %dopar% compScoreFunc(plat)
  } else {
    estimatedScoreMat <- foreach(plat = 1:length(platform_list), .combine='cbind') %do% compScoreFunc(plat)
  }
	
	########################################
	# if necessary, add permuted score for each platform as well as integrated data
	########################################
  if(addPermutedScore==TRUE){
    if(echo==TRUE){
      cat('Calculating permuted latent trait for each platform', '\n')	
    }	
    compPermutedScoreFunc <- function(i) {
      calculatePermutedScoreByGeneSampling(platform_list[[i]], 
                                           dscrmn=a_list[[i]],
                                           dffclt=b_list[[i]], 
                                           c=c_list[[i]], fold=fold)
    }
    plat <- NULL # to quite R CMD check
    if(parallel==TRUE){
      permutedScoreMat <- foreach(plat = 1:length(platform_list), .combine='cbind', .packages='integIRTy') %dopar% compPermutedScoreFunc(plat)
    } else {
      permutedScoreMat <- foreach(plat = 1:length(platform_list), .combine='cbind') %do% compPermutedScoreFunc(plat)
    }
		
  }
	# returns integrated latent trait and ltm fit on individual platform					
  if(addPermutedScore==TRUE){
    return(list(fits=fits, estimatedScoreMat=estimatedScoreMat, permutedScoreMat=permutedScoreMat, dscrmnList=a_list, dffcltList=b_list, gussngList=c_list))
  } else {
    return(list(fits=fits, estimatedScoreMat=estimatedScoreMat, dscrmnList=a_list, dffcltList=b_list, gussngList=c_list))
  }
}