R/fitOnSinglePlat.R

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

fitOnSinglePlat <-
function(data, model=2, guessing=FALSE, 
         sampleIndices=1:ncol(data), 
         geneIndices=1:nrow(data), ...) {
	## tpm fits a guessing parameter in 2 modes: one is guessing in unconstrained rasch,
	## the other is guessing in 2PL.
	## if guessing=T, then max.guessing=1 is used to fit the guessing parameter for every item
	#K <- length(sampleIndices) # item size
  if(is.null(sampleIndices)) sampleIndices=1:ncol(data)
  if(is.null(geneIndices)) geneIndices=1:nrow(data)
	
  temp <- data[geneIndices, sampleIndices]
  if(model==1) {
    tempFit <- try(rasch(temp, constraint = cbind(ncol(temp) + 1, 1), ...),
                   silent=TRUE)
  }
  if(model==2) {
  # force all dscrmn the same. dscmn is contained as a pure vector in the coef.
    if(guessing==FALSE) { # no guessing
      tempFit <- try(rasch(temp, ...), silent=TRUE)
    } else { # with guessing
      tempFit <- try(tpm(temp, type = "rasch", ...) , silent=TRUE)
    }
  }
  if(model==3) {
    if(guessing==FALSE) {# no guessing
      tempFit <- try(ltm(temp ~ z1, ...), silent=TRUE)
    } else { # with guessing
      tempFit <- try(tpm(temp, ...) , silent=TRUE)
    } 	
  }
  return(list(fit=tempFit, model=model, guessing=guessing,
              sampleIndices=sampleIndices, geneIndices=geneIndices))
}