R/reglca_init_parameters.R

In CDM: Cognitive Diagnosis Modeling

## File Name: reglca_init_parameters.R
## File Version: 0.168

reglca_init_parameters <- function( nclasses, dat0, sd_noise_init, item_probs_init,
        class_probs_init, random_starts, G, est_type=NULL)
{
    if (is.null(est_type)){
        est_type <- "CD"
    }
    use_random_starts <- FALSE
    means <- colMeans(dat0, na.rm=TRUE )
    I <- ncol(dat0)
    #--- initial class probabilities
    if ( is.null(class_probs_init) ){
        class_probs <- reglca_init_parameters_class_probs( nclasses=nclasses,
                            sd_noise_init=sd_noise_init, G=G)
    } else {
        class_probs <- class_probs_init
        random_starts <- 1
    }

    #--- item probabilities
    qmeans <- stats::qnorm(means)
    if ( is.null(item_probs_init) ){
        item_probs <- reglca_init_parameters_item_probs( qmeans=qmeans, I=I, nclasses=nclasses,
                                sd_noise_init=sd_noise_init, parm_range=1 )
    } else {
        item_probs <- item_probs_init
        random_starts <- 1
    }
    #- transformation into logit parameters
    xsi <- NULL
    if (est_type=="DIFF"){
        xsi <- stats::qlogis(p=item_probs)
    }

    #--- item probabilities in case of random starts
    if (random_starts > 1){
        item_probs <- list()
        sd_noise_init <- max( sd_noise_init, .01 )
        for (rr in 1:random_starts){
            item_probs[[rr]] <- reglca_init_parameters_item_probs( qmeans=qmeans,
                            I=I, nclasses=nclasses, sd_noise_init=sd_noise_init,
                            parm_range=1 )
        }
    }

    #--- class probabilities in case of random starts
    if (random_starts > 1){
        class_probs <- list()
        for (rr in 1:random_starts){
            class_probs[[rr]] <- reglca_init_parameters_class_probs( nclasses=nclasses,
                                        sd_noise_init=sd_noise_init, G=G)
        }
        use_random_starts <- TRUE
    }

    #--- output
    res <- list( class_probs=class_probs, item_probs=item_probs, random_starts=random_starts,
                        use_random_starts=use_random_starts, est_type=est_type,
                        xsi=xsi )
    return(res)
}