R/setup.data.R

In irtpwr: Power Analysis for IRT Models Using the Wald, LR, Score, and Gradient Statistics

setup.data <- function(hyp, n, dist.fun = stats::rnorm) {
    # Create artificial dataset from alternative
    # hypothesis @param hyp Hypothesis Object
    # created from the setup.hypothesis function
    # @param n integer, number of persons @param
    # dist.fun function to generate the person
    # parameters @return @export @examples
    # altpars <- list( a = rlnorm(5,sdlog = .4),
    # d = stats::rnorm(5)) hyp <-
    # setup.hypothesis(type = '1PLvs2PL', altpars
    # = altpars) data <-
    # setup.data(hyp=hyp,n=500)

    distfun <- function(x) matrix(dist.fun(x), ncol = 1)

    if (isTRUE(hyp$unresmod$multigroup)) {
        # Multigroup Model

        df <- lapply(hyp$unresmod$parsets, function(pars) mirt::simdata(a = pars$a,
            d = pars$d, Theta = distfun(n/2), itemtype = hyp$unresmod$itemtype)) |>
            (function(x) do.call(rbind, x))()
        group <- rep(c("A", "B"), each = n/2)
        re <- list(data = df, group = group)

    } else {
        # SingleGroup Model

        pars <- hyp$unresmod$parsets
        is.multi <- "a2" %in% colnames(pars$a)
        if (is.multi) {
            distfun <- function(x) {
                matrix(dist.fun(2 * x), ncol = 2)
            }
        }

        if (is.null(pars$g)) {
            pars$g <- 0
        }
        df <- mirt::simdata(a = pars$a, d = pars$d,
            guess = pars$g, Theta = distfun(n), itemtype = hyp$unresmod$itemtype)
        re <- list(data = df)
    }

    return(re)
}