R/iccdeboot.R

In iccde: Computation of the Double-Entry Intraclass Correlation

#' Computation of Double-Entry Intraclass Correlation
#'
#' Test for matrices comprising of correlations
#'
#' @param data A data frame with participants in rows and variables in columns.
#' Users should restrict the data set to the variables of concrete interest
#' because the all available information in the data frame will be used to
#' compute bootstrapped confidence intervals. Thus, it is advisable to create a
#' new data frame that entails only the variables on which the matrix of ICCDEs
#' should be based.
#' @param n.sim The number of iterations to be carried out. Default is 1,000.
#' @param alpha Type I error. Default is .05.
#' @param digit Number of digits in the output. Default is 3.
#'
#' @return The output provides a list of matrices of double-entry intraclass
#' correlations. The first matrix comprises the bootstrapped point estimates of
#' the bootstrapped double-entry intraclass correlations, the second contains
#' the lower limits of the bootstrap confidence intervals, given the desired
#' alpha level, the third one contains the upper limits of the bootstrap
#' confidence intervals, given the desired alpha level.
#'
#' @export
#'
#' @examples
#' df <- data.frame(a = rnorm(100), b = rnorm(100), c = rnorm(100),
#'                  x = rnorm(100), y = rnorm(100), z = rnorm(100))
#' icc.de.boot(data = df,
#'             n.sim = 1000,
#'             alpha = .01,
#'             digit = 2)

icc.de.boot <- function(data,
                        n.sim = 1000,
                        alpha = .05,
                        use = "pairwise",
                        digit = 3){

  array <- array(0, dim = c(ncol(data),
                            ncol(data),
                            n.sim))

  for(m in 1:n.sim){
    index <- sample(1:nrow(data), replace = TRUE)

    cormat <- cor(data[index, ],
                  use = use)

    r <- matrix(0, nrow = ncol(cormat), ncol = ncol(cormat))
    for (i in 1:nrow(cormat)){
      for (j in 1:ncol(cormat)){
        r[i, j] <- icc.de(cormat[-c(i, j), i], cormat[-c(i, j), j],
                          digits = digit)
      }
    }

    colnames(r) <- colnames(data)
    rownames(r) <- colnames(data)

    array[1:nrow(cormat), 1:ncol(cormat), m] <- r
  }

  dimnames(array) <- list(colnames(r), colnames(r), 1:n.sim)

  out <- list(Mean = round(apply(array, c(1, 2),
                                 function(x) mean(x,
                                                  na.rm = TRUE)),
                           digit),
              LL = round(apply(array, c(1, 2),
                               function(x) quantile(x,
                                                    probs = (alpha / 2),
                                                    na.rm = TRUE)),
                         digit),
              UL = round(apply(array, c(1, 2),
                               function(x) quantile(x,
                                                    probs = 1 - (alpha / 2),
                                                    na.rm = TRUE)),
                         digit))

  return(out)
}