R/item_by_group.R

In CDM: Cognitive Diagnosis Modeling

## File Name: item_by_group.R
## File Version: 0.19

##########################################################
# creates an extended dataset with item responses in which
# items are defined as combinations of original items and group
item_by_group <- function( dat, group, invariant=NULL, rm.empty=TRUE )
{
    vars <- colnames(dat)
    some_invariant_items <- ( ! is.null(invariant) )
    if ( some_invariant_items ){
        vars <- setdiff(vars,invariant)
    }

    I <- length(vars)
    group_unique <- sort( unique(group) )
    G <- length(group_unique)
    #*** create extended dataset
    dat2 <- matrix( NA, nrow=nrow(dat), ncol=I*G )
    cn <- sapply( vars, FUN=function(vv){
                paste0( vv, "_group", group_unique ) }, simplify=FALSE)
    colnames(dat2) <- unlist(cn)
    for (gg in 1:G){
        # gg <- 1
        ind_gg <- which( group==group_unique[gg] )
        for (ii in 1:I){
            # ii <- 1
            dat2[ ind_gg, G*(ii-1) + gg ] <- dat[ ind_gg, vars[ii] ]
        }
    }
    #--- include invariant items
    if ( some_invariant_items ){
        dat2a <- dat[, invariant]
        dat2 <- cbind( dat2a, dat2 )
    }

    #--- remove empty columns
    if (rm.empty){
        ind <- which( colMeans( is.na(dat2) )==1 )
        if ( length(ind) > 0 ){
            dat2 <- dat2[, - ind ]
        }
    }
    #--- include some attributes: variables and variable indices
    attr(dat2,"noninvariant") <- vars
    attr(dat2,"invariant") <- invariant
    attr(dat2,"noninvariant_index") <- match( vars, colnames(dat))
    attr(dat2,"noninvariant_index_extended") <-
            rep( attr(dat2,"noninvariant_index"), each=G )
    attr(dat2,"invariant_index") <- match( invariant, colnames(dat))
    attr(dat2,"all_index") <- c( attr(dat2,"invariant_index"), attr(dat2,"noninvariant_index_extended") )
    cn <- colnames(dat2)
    names(cn) <- NULL
    colnames(dat2) <- cn
    #--- output
    return(dat2)
}
############################################################