R/isop.scoring.R

Defines functions isop.scoring

Documented in isop.scoring

## File Name: isop.scoring.R
## File Version: 1.11
#################################################
# scoring students and items according to the
# Scheiblechner's ISOP model
isop.scoring <- function( dat, score.itemcat=NULL){
    #***
    # compute modified percentile score
    maxK <- apply( dat, 2, max, na.rm=TRUE )
    K <- max(maxK)
    I <- ncol(dat)
    # define response matrix
    dat.resp <- 1-is.na( dat )
    # p values for every category
    p.itemcat <- matrix(NA, nrow=I, ncol=K+1 )
    rownames(p.itemcat) <- colnames(dat)
    colnames(p.itemcat) <- paste0( "p.Cat", 0:K )
    for (kk in 0:K){
        p.itemcat[,kk+1] <- colMeans( dat==kk, na.rm=TRUE )
                }
    #**************************************
    if ( is.null( score.itemcat )){
        score.itemcat <- 0*p.itemcat
        colnames(score.itemcat) <- paste0( "score.Cat", 0:K )
        # score itemcat
        for (kk in 0:K){
            if (kk>0){
                v1 <- apply( p.itemcat[, 1+seq( 0, kk -1, 1 ), drop=FALSE],
                            1, sum, na.rm=TRUE )
                    } else { v1 <- 0 }
            if (kk<K){
                v2 <- apply( p.itemcat[, 1+seq( kk+1, K, 1 ), drop=FALSE],
                            1, sum, na.rm=TRUE )
                    } else { v2 <- 0 }
            score.itemcat[,kk+1] <- v1-v2
                    }
                }  # end score itemcat
      #*************************
      dat <- dat[, rownames(score.itemcat) ]
      # scoring table
      score.dat <- 0*dat
      for (kk in 0:K){
        score.dat <- score.dat + matrix( score.itemcat[,kk+1], nrow=nrow(dat),
                        ncol=I, byrow=TRUE ) * ( dat==kk )
                    }
      # score persons
      person <- data.frame( "pid"=1:nrow(dat),
                    "score"=rowSums( dat, na.rm=TRUE ),
                    "max"=rowSums( dat.resp ),
                    "M"=rowMeans( dat, na.rm=TRUE ),
                    "mpsc"=rowMeans( score.dat, na.rm=TRUE )
                            )
      # compute item p score
      item.pscore <- rep(NA,I)
      for (ii in 1:I){
#          ii <- 1
          v1 <- colMeans( dat[,ii ] > dat[,-ii] )
          v2 <- colMeans( dat[,ii ] < dat[,-ii] )
          item.pscore[ii] <- mean( v1 - v2 )
                        }
      # score items
      item <- data.frame( "item"=colnames(dat),
               "M"=colMeans( dat, na.rm=TRUE ) )
      item$pscore <- item.pscore
      item <- cbind( item, p.itemcat,  score.itemcat )
      # calculate distribution function
      distr.fct <- t(apply( p.itemcat, 1, FUN=function(ll){
                    cumsum( ll ) } ) )
#      colnames(item)[-c(1:2)] <- colnames( score.itemcat )[-1]
      #***
      # collect results
      res <- list( "person"=person, "item"=item,
            "p.itemcat"=p.itemcat, "score.itemcat"=score.itemcat,
            "distr.fct"=distr.fct)
      return(res)
      }
##############################################################
# example scoring for item with 3 categories
# probabilities p0, p1, p2
# score category 0: s0=- ( p1 + p2 )
# score category 1: s1=p0 - p2
# score category 2: s2=p0 + p1
# ***
# s2 + s0=p0 + p1 - p1 - p2=p0 - p2=s1
# ***
# s2 - s1=p0 + p1 - p0 + p2=p1 + p2
# s1 - s0=p0 - p2 + p1 + p2=p0 + p1
alexanderrobitzsch/sirt documentation built on March 18, 2024, 1:29 p.m.