R/scaledlogit.R

In JobNmadu/Dyn4cast: Dynamic Modeling and Machine Learning Environment

#' Scale Parameter for Integer Modeling and Forecast
#'
#' @description
#' This function is a wrapper for scaling the fitted (predicted) values of a
#' one-sided (positive or negative only) integer response variable of supported
#'  models. The scaling involves some log transformation of the fitted
#'  (predicted) values.
#'
#' @param x The parameter to be scaled, which is the fitted values from
#' supported models. The scaled parameter is used mainly for constrained
#' forecasting of a response variable _positive (0 - inf) or negative
#' (-inf - 0)_. The scaling involves log transformation of the parameter
#' @param lower Integer or variable representing the lower limit for the
#' scaling (-inf or 0)
#' @param upper Integer or variable representing the upper limit for the
#' scaling (0 or inf)
#'
#' @export scaledlogit
#' @name scaledlogit
#'
#' @examples
#' library(Dyn4cast)
#' library(splines)
#' lower <- 1
#' upper <- 37
#' Model   <- lm(states ~ bs(sequence, knots = c(30, 115)), data = Data)
#' scaledlogit(x = fitted.values(Model), lower = lower,
#'  upper = upper)
scaledlogit <- function(x, lower, upper) {
  log((x - lower) / (upper - x))
}