tests/testthat/helper-macro-micro.R

In yardstick: Tidy Characterizations of Model Performance

# These helpers are used to test macro and macro weighted methods

data_three_by_three <- function() {
  as.table(
    matrix(
      c(
        3, 1, 1,
        0, 4, 2,
        1, 3, 5
      ),
      ncol = 3, byrow = TRUE,
      dimnames = list(c("c1", "c2", "c3"), c("c1", "c2", "c3"))
    )
  )
}

multi_ex <- data_three_by_three()
weighted_macro_weights <- colSums(multi_ex) / sum(colSums(multi_ex))

# turn a 3x3 conf mat into a 2x2 submatrix in a one vs all approach
make_submat <- function(data, col) {
  top_left <- data[col, col]
  top_righ <- sum(data[col, -col])
  bot_left <- sum(data[-col, col])
  bot_righ <- sum(data[-col, -col])
  as.table(
    matrix(
      c(top_left, top_righ, bot_left, bot_righ),
      ncol = 2,
      byrow = TRUE
    )
  )
}

# These are the "one vs all" sub matrices
# for macro / weighted macro, calculate the binary version of each metric
# and then average them together
multi_submats <- list(
  c1 = make_submat(multi_ex, 1),
  c2 = make_submat(multi_ex, 2),
  c3 = make_submat(multi_ex, 3)
)

# Just pass in a binary metric function
macro_metric <- function(binary_metric, event_level = "first", ...) {
  mean(
    vapply(multi_submats, binary_metric, numeric(1), event_level = event_level, ...)
  )
}

macro_weighted_metric <- function(binary_metric, event_level = "first", ...) {
  stats::weighted.mean(
    vapply(multi_submats, binary_metric, numeric(1), event_level = event_level, ...),
    weighted_macro_weights
  )
}

# For micro examples, we calculate the pieces by hand and use them individually
data_three_by_three_micro <- function() {
  res <- list(
    tp = vapply(multi_submats, function(x) {
      x[1, 1]
    }, double(1)),
    p = vapply(multi_submats, function(x) {
      colSums(x)[1]
    }, double(1)),
    tn = vapply(multi_submats, function(x) {
      x[2, 2]
    }, double(1)),
    n = vapply(multi_submats, function(x) {
      colSums(x)[2]
    }, double(1))
  )

  res <- c(
    res,
    list(
      fp = res$p - res$tp,
      fn = res$n - res$tn
    )
  )

  res
}