R/sim.gdina.R

In CDM: Cognitive Diagnosis Modeling

## File Name: sim.gdina.R
## File Version: 2.20



################################################################################
# Simulation of the GDINA model
sim.gdina <- function( n, q.matrix, delta, link="identity",
                thresh.alpha=NULL, cov.alpha=NULL, alpha=NULL,
                Mj, Aj, necc.attr )
{
    I <- length(delta)
    # simulate alpha
    if ( is.null(alpha) ){
        alpha <- 1* ( CDM_rmvnorm( n, mean=thresh.alpha, sigma=cov.alpha ) > 0 )
    }
    dat <- matrix( NA, n, I )
    for (ii in 1:I){
        na.ii <- necc.attr[[ii]]
        Aj.ii <- Aj[[ii]]
        Mj.ii <- Mj[[ii]][[1]]
        delta.ii <- delta[[ii]]
        alpha.ii <- alpha[, na.ii, drop=FALSE ]
        # calculate probability for every attribute pattern
        patt.prob <- rowSums( Mj.ii * outer( rep(1,nrow(Mj.ii) ), delta.ii  ) )
        # create patterns for alpha.ii and Aj.ii
        l1.Aj <- l1.al <- "P"
        for (vv in 1:( ncol(alpha.ii) ) ){
            l1.Aj <- paste( l1.Aj, Aj.ii[,vv], sep="")
            l1.al <- paste( l1.al, alpha.ii[,vv], sep="")
        }
        resp.ii <- patt.prob[ match( l1.al, l1.Aj ) ]
        if ( link=="logit"){
            resp.ii <- stats::plogis( resp.ii )
        }
        if ( link=="log"){
            resp.ii <- exp( resp.ii )
        }
        dat[, ii] <- 1*( stats::runif(n) < resp.ii )
    }
    res <- list( data=dat, alpha=alpha, q.matrix=q.matrix, delta=delta,
                        Aj=Aj, Mj=Mj, link=link )
    return(res)
}