R/lambda_inf.R

In softloud/lambda: What the Package Does (One Line, Title Case)

#' Choose a threshold to classify observations that are consistent with theory m
#'
#' To determine which observations are consistent with each theory or none,
#' a threshold must be chosen above which an observation \eqn{i} is considered
#' consistent with theory \eqn{m}. This function considers all theories and
#' observations and returns the largest threhold.
#'
#' The function calculates the maximum threshold \eqn{\lambda} for which
#' (probability of not happening)/(probability of happening) < \eqn{\alpha}
#' for values that exceed threshold \eqn{\lambda}.
#' (Defaults to \eqn{\alpha} := 0.05)
#'
#' \eqn{\lambda} is a possible value for the threshold for an odds-ratio of
#' posterior probabilities that are greater than alpha, where the posterior
#' probability \eqn{p_{im}} of observation \eqn{i} for theory \eqn{m} exceed the
#' threshold. The previous sentence irrefutably demonstrates why we \emph{need}
#' mathematics. Anyways, check out the equation, let me know if I borked the
#' maths.
#'
#' @inheritParams lambda_fn
#' @param precision what level of precision is required
#' @export

lambda_inf <- function(probability, alpha = 0.5, precision = 0.001) {
  # optimise(
  #   lambda_fn,
  #   interval = c(0, 1),
  #   prob = probability,
  #   alpha = alpha,
  #   maximum = FALSE
  # ) %>%
  #   purrr::pluck("minimum")

  tibble(
    lambda = seq(0, 1, by = 0.001),
    lambda_fn = map_dbl(lambda, lambda:::lambda_fn, probability)
  ) %>% pluck("lambda_fn") %>% min()

  }