R/vcov.loglike.din.R

In CDM: Cognitive Diagnosis Modeling

## File Name: vcov.loglike.din.R
## File Version: 0.10


#########################################################################
# compute log-likelihood for din objects
vcov.loglike.din <- function( weights, skillprobs0, slip0, guess0,
        latresp, item.patt.split, resp.ind.list,
        return.p.xi.aj=FALSE )
{
    ########################
    IP <- N <- length(weights)
    L <- length(skillprobs0)
    J <- length(guess0)
    # calculate probabilities
    slipM <- matrix( slip0, nrow=nrow(latresp), ncol=ncol(latresp))
    guessM <- matrix( guess0, nrow=nrow(latresp), ncol=ncol(latresp))
    pj <- (1 - slipM )*latresp + guessM * ( 1 - latresp )
    pjM <- array( NA, dim=c(J,2,L) )
    pjM[,1,] <- 1 - pj
    pjM[,2,] <- pj
    skillprobsM <- matrix( skillprobs0, nrow=IP, ncol=L, byrow=TRUE )
    # calculate log-likelihood
    h1 <- matrix( 1, nrow=IP, ncol=L )
    res.hwt <- cdm_calc_posterior(rprobs=pjM, gwt=h1, resp=item.patt.split,
                        nitems=J, resp.ind.list=resp.ind.list, normalization=FALSE,
                        thetasamp.density=NULL, snodes=0 )
    p.xi.aj <- res.hwt$hwt
    # Log-Likelihood (casewise)
    ll2 <- log( rowSums( p.xi.aj * skillprobsM ) )
    if (return.p.xi.aj){
        res <- list( "ll"=ll2, "p.xi.aj"=p.xi.aj )
    }  else {
        res <- ll2
    }
    return(res)
}
#########################################################################