R/gdina.wald_aux.R

In alexanderrobitzsch/CDM: Cognitive Diagnosis Modeling

## File Name: gdina.wald_aux.R
## File Version: 0.18

##############################################
##############################################
# calculate restricted model and output
# distance to estimated GDINA model
calc_dist_restricted_model <- function(pjj, Mj.ii0, attr.prob.ii, link,
        suffstat_probs.ii, aggr.attr.patt.ii  )
{
    mod0.ii <- stats::lm( pjj ~ 0 + Mj.ii0, weights=attr.prob.ii )
    fitted.ii <- stats::fitted(mod0.ii)
    if ( link=="logit"){ fitted.ii <- stats::plogis(fitted.ii) }
    if ( link=="log"){ fitted.ii <- exp(fitted.ii) }
    # weighted distance
    dist.ii <- sum( attr.prob.ii * ( suffstat_probs.ii - fitted.ii)^2 )
    # unweighted distance
    d1 <- ( suffstat_probs.ii - fitted.ii)^2
    d1 <- as.vector(d1[ aggr.attr.patt.ii ])
    d1 <- mean( d1 )
    res <- list( "wgtdist"=dist.ii, "uwgtdist"=d1)
    return(res)
}
###################################################

###################################################
# creation of constraint matrix
contraint_matrix <- function( delta.ii, rule, Kii, Mj.ii )
{
    Dii <- length(delta.ii)
    #***************************
    # DINA
    if ( rule=="DINA"){
        R <- matrix( 0, nrow=Dii-2, ncol=Dii)
        for (vv in 2:(Dii-1) ){
            R[vv-1,vv] <- 1
        }
    }
    #***************************
    # ACDM
    if (rule=="ACDM"){
        R <- matrix( 0, nrow=Dii-(Kii+1), ncol=Dii)
        for (vv in 1:(nrow(R)) ){
            vv1 <- vv + ( Kii +1 )
            R[vv,vv1] <- 1
        }
    }
    #***************************
    # DINO
    if (rule=="DINO"){
        Dii <- ncol(Mj.ii)
        R <- matrix( 0, nrow=Dii-2, ncol=Dii)
        for (vv in 1:(Dii-2)){
            R[vv,] <- Mj.ii[vv+2,] - Mj.ii[vv+1, ]
        }
    }
    #**********************************
    return(R)
}
####################################################