R/crosstab.R

In fwildclusterboot: Fast Wild Cluster Bootstrap Inference for Linear Models

# crosstab2 <- function(data, var1, var2) {
#
#   #' Function 2 to calculate crosstabs
#   #' @param data A matrix to collapse by two dimensions var1 var2
#   #' @param var1 a data.frame containing a single variable
#   #' @param var2 a data.frame containing a single variable
#   #' @return A collapsed matrix of dimension length(unique(var1)) x
# length(unique(var2)). If...
#
#   # data = XinvXXr
#   # var1 = clustid
#   # var2 = fixed_effect
#   dreamerr::check_arg(var1, "data.frame")
#   dreamerr::check_arg(var2, "data.frame")
#
#   dt <- data.table(var1 = var1, var2 = var2, y = data)
#   setnames(dt, names(dt), c("var1", "var2", "y"))
#   setkeyv(dt, c("var1", "var2"))
#
#   dt <- dt[CJ(var1, var2, unique = TRUE), lapply(.SD, sum), by = .EACHI]
#   dt[is.na(y), y := 0]
#
#   unique_var1 <- as.vector(unique(var1[, names(var1)]))
#   # unique_var2 <- unique(var2)
#   unique_var2 <- as.vector(unique(var2[, names(var2)]))
#
#
#   unique_var1_len <- length(unique_var1)
#   unique_var2_len <- length(unique_var2)
#
#   # dt$y
#   matrix(dt$y, unique_var1_len, unique_var2_len, byrow = TRUE)
# }


#' Function 4 to calculate crosstabs
#' @param data A matrix to collapse by two dimensions var1 var2
#' @param var1 a data.frame containing a single variable
#' @param var2 a data.frame containing a single variable
#' @return A collapsed matrix of dimension length(unique(var1)) x
#' length(unique(var2)). If...
#' @importFrom stats aggregate
#' @importFrom dreamerr check_arg
#' @noRd

crosstab4 <- function(data, var1, var2) {
  dreamerr::check_arg(var1, "data.frame")
  dreamerr::check_arg(var2, "data.frame")


  length_var1 <- nrow(unique(var1))
  length_var2 <- nrow(unique(var2))
  res <-
    stats::aggregate(data, data.frame(var1, var2), sum, drop = FALSE)
  # res <- Matrix.utils::aggregate.Matrix(data, list(var1, var2),
  #         sum, drop = FALSE)
  res <- res[, 3]
  res[is.na(res)] <- 0
  dim(res) <- c(length_var1, length_var2)
  res
}

# #' collapse way to calculate crosstabs
# #' @param data A matrix to collapse by two dimensions var1 var2
# #' @param var1 a data.frame containing a single variable
# #' @param var2 a data.frame containing a single variable
# #' @return A collapsed matrix of dimension length(unique(var1)) x
# # length(unique(var2)). If...
# #' @importFrom dreamerr check_arg
# #' @importFrom collapse fsum qF
# #' @noRd
#
# crosstab3 <- function(data, var1, var2){
#
#   dreamerr::check_arg(var1, "data.frame")
#   dreamerr::check_arg(var2, "data.frame")
#
#   length_var1 <- nrow(unique(var1))
#   length_var2 <- nrow(unique(var2))
#
#   res <-
#     collapse::fsum(
# x = data, g = collapse::qF(var1[, 1]):collapse::qF(var2[, 1]))
#
#   res <- matrix(res, length_var1, length_var2, byrow = TRUE)
#   res[is.na(res)] <- 0
#   res
# }

#' optimized collapse way to calculate crosstabs
#' @param data A matrix to collapse by two dimensions var1 var2
#' @param var1 a data.frame containing a single variable
#' @param var2 a data.frame containing a single variable
#' @return A collapsed matrix of dimension length(unique(var1)) x
#' length(unique(var2)). If...
#' @importFrom dreamerr check_arg
#' @importFrom collapse fsum qF
#' @noRd

crosstab <- function(data, var1, var2) {
  dreamerr::check_arg(var1, "data.frame")
  dreamerr::check_arg(var2, "data.frame")

  f1 <- collapse::qF(.subset2(var1, 1L), na.exclude = FALSE)
  f2 <- collapse::qF(.subset2(var2, 1L), na.exclude = FALSE)
  intact <- f1:f2
  class(intact) <- c(class(intact), "na.included")
  # See https://sebkrantz.github.io/collapse/articles/
  # collapse_intro.html#factors-grouping-objects-and-grouped-data-frames
  res <- collapse::fsum(data, intact, use.g.names = FALSE)
  lev1 <- attr(f1, "levels")
  lev2 <- attr(f2, "levels")
  res <- matrix(
    res,
    length(lev1),
    length(lev2),
    byrow = TRUE,
    dimnames = list(lev1, lev2)
  )
  res[is.na(res)] <- 0
  res
}