Add list columns of p/d/q/r functions to a forecast data frame

Given a data frame with one row per target, a dist column (or a dist argument) naming a distribution, and columns for that distribution's parameters, adds list columns holding the corresponding cdf, density, quantile function and/or random generator. allocate() needs only the cdf ("p") and quantile function ("q").

Usage

add_pdqr_funs(
  df,
  dist = df$dist,
  trans = NULL,
  trans_inv = NULL,
  transpars = NULL,
  types = c("p", "d", "q", "r"),
  fnames = c(p = "F", d = "f", q = "Q", r = "r")
)

Arguments

df

data frame with one row per target. It must have a column for each parameter required by the distribution functions named in a dist column or in the dist argument.

dist

character vector of length 1 or nrow(df) giving distribution root names (or "distfromq"); required if df has no dist column.

trans

function of an outcome x and parameters, pre-composed with the p and d functions. Use this for distributions whose support must be rescaled, such as a beta distribution on [x_min, x_max].

trans_inv

inverse of trans, post-composed with the q and r functions.

transpars

parameters for trans and trans_inv that are not columns of df.

types

which of "p", "d", "q", "r" to add.

fnames

names to give the added list columns.

Value

A tibble with the requested list columns added.

Examples

fc <- add_pdqr_funs(
  tibble::tibble(
    target_names = c("a", "b"),
    dist = "norm",
    mean = c(5, 8),
    sd = c(1, 2)
  ),
  types = c("p", "q")
)
fc$F[[1]](5)
#> [1] 0.5