R/gdina_post_pattern_output.R

In CDM: Cognitive Diagnosis Modeling

## File Name: gdina_post_pattern_output.R
## File Version: 0.07

gdina_post_pattern_output <- function(G, p.xi.aj, zeroprob.skillclasses,
    item.patt, attr.patt.c, p.aj.xi, item.patt.subj, group2, attr.patt, K )
{

    # calculate posterior probability for each attribute pattern
    if (G==1){
        # set likelihood for skill classes with zero probability to zero
        if ( ! is.null(zeroprob.skillclasses) ){
            p.xi.aj[, zeroprob.skillclasses ] <- 0
        }
        pattern <- data.frame(
                        freq=round(as.numeric(item.patt[,-1]),3),
                        mle.est=attr.patt.c[ max.col( p.xi.aj ) ],
                        mle.post=rowMaxs( p.xi.aj ) / rowSums( p.xi.aj ),
                        map.est=attr.patt.c[ max.col( p.aj.xi ) ],
                        map.post=rowMaxs( p.aj.xi ) )
    }
    if (G>1){
        ind1 <- match( item.patt.subj, item.patt[,1] )
        l1 <- attr.patt.c[ max.col( p.xi.aj ) ]
        pattern <- data.frame( "mle.est"=l1[ind1] )
        l1 <- rowMaxs( p.xi.aj ) / rowSums( p.xi.aj )
        pattern$mle.post <- l1[ind1]
        pattern$map.est <- NA
        pattern$map.post <- NA
        for (gg in 1:G){
            # gg <- 1
            ind.gg <- which( group2==gg )
            ind2.gg <- match( item.patt.subj[ind.gg], item.patt[, 1] )
            l1 <- attr.patt.c[ max.col( p.aj.xi[,,gg] ) ]
            pattern$map.est[ind.gg] <- l1[ind2.gg]
            l1 <-  rowMaxs( p.aj.xi[,,gg] )
            pattern$map.post[ind.gg] <- l1[ind2.gg]
                    }
            }
    # calculate posterior probabilities for all skills separately
    if (G==1){
        attr.postprob <- p.aj.xi %*% attr.patt
        colnames( attr.postprob ) <- paste("post.attr",1:K, sep="")
        pattern <- cbind( pattern,  attr.postprob )
    }
    #--------- OUTPUT
    res <- list( pattern=pattern, p.xi.aj=p.xi.aj )
    return(res)
}