diff --git a/DESCRIPTION b/DESCRIPTION
index 4e59523..08959d6 100644
--- a/DESCRIPTION
+++ b/DESCRIPTION
@@ -1,6 +1,6 @@
Package: probably
Title: Tools for Post-Processing Predicted Values
-Version: 1.1.1.9000
+Version: 1.1.1.9001
Authors@R: c(
person("Max", "Kuhn", , "max@posit.co", role = c("aut", "cre")),
person("Davis", "Vaughan", , "davis@posit.co", role = "aut"),
diff --git a/NAMESPACE b/NAMESPACE
index 23ed2dd..de28cd6 100644
--- a/NAMESPACE
+++ b/NAMESPACE
@@ -126,6 +126,10 @@ S3method(required_pkgs,cal_estimate_multinomial)
S3method(required_pkgs,cal_estimate_multinomial_spline)
S3method(required_pkgs,cal_estimate_none)
S3method(required_pkgs,cal_object)
+S3method(required_pkgs,int_conformal_cv)
+S3method(required_pkgs,int_conformal_full)
+S3method(required_pkgs,int_conformal_quantile)
+S3method(required_pkgs,int_conformal_split)
S3method(threshold_perf,data.frame)
S3method(type_sum,cal_object)
S3method(vec_cast,character.class_pred)
diff --git a/NEWS.md b/NEWS.md
index e788116..7be95a0 100644
--- a/NEWS.md
+++ b/NEWS.md
@@ -1,5 +1,7 @@
# probably (development version)
+* Add `required_pkgs()` methods to `int_conformal_cv()`, `int_conformal_full()`, `int_conformal_quantile()`, and `int_conformal_split()`. (#190)
+
# probably 1.1.1
* Updated unit tests for new ggplot2 release (#180).
diff --git a/R/conformal_infer.R b/R/conformal_infer.R
index 10fadfa..6511193 100644
--- a/R/conformal_infer.R
+++ b/R/conformal_infer.R
@@ -113,6 +113,20 @@ print.int_conformal_full <- function(x, ...) {
invisible(x)
}
+#' @rdname required_pkgs.int_conformal_cv
+#' @keywords internal
+#' @export
+required_pkgs.int_conformal_full <- function(x, infra = TRUE, ...) {
+ model_pkgs <- required_pkgs(x$wflow, infra = infra)
+
+ if (infra) {
+ model_pkgs <- c(model_pkgs, "probably")
+ }
+
+ model_pkgs <- unique(model_pkgs)
+ model_pkgs
+}
+
# ------------------------------------------------------------------------------
#' Prediction intervals from conformal methods
diff --git a/R/conformal_infer_cv.R b/R/conformal_infer_cv.R
index 41a774c..79c2ebb 100644
--- a/R/conformal_infer_cv.R
+++ b/R/conformal_infer_cv.R
@@ -148,6 +148,23 @@ print.int_conformal_cv <- function(x, ...) {
invisible(x)
}
+#' S3 methods to track which additional packages are needed for prediction
+#' intervals via conformal inference
+#' @param x a conformal interval object
+#' @inheritParams generics::required_pkgs
+#' @export
+required_pkgs.int_conformal_cv <- function(x, infra = TRUE, ...) {
+ model_pkgs <- map(x$models, required_pkgs, infra = infra)
+ model_pkgs <- unlist(model_pkgs)
+
+ if (infra) {
+ model_pkgs <- c(model_pkgs, "probably")
+ }
+
+ model_pkgs <- unique(model_pkgs)
+ model_pkgs
+}
+
# ------------------------------------------------------------------------------
# helpers
diff --git a/R/conformal_infer_quantile.R b/R/conformal_infer_quantile.R
index c2227b5..3f7803c 100644
--- a/R/conformal_infer_quantile.R
+++ b/R/conformal_infer_quantile.R
@@ -115,6 +115,20 @@ print.int_conformal_quantile <- function(x, ...) {
invisible(x)
}
+#' @rdname required_pkgs.int_conformal_cv
+#' @keywords internal
+#' @export
+required_pkgs.int_conformal_quantile <- function(x, infra = TRUE, ...) {
+ model_pkgs <- required_pkgs(x$wflow, infra = infra)
+
+ if (infra) {
+ model_pkgs <- c(model_pkgs, "probably")
+ }
+
+ model_pkgs <- unique(model_pkgs)
+ model_pkgs
+}
+
#' @export
#' @rdname predict.int_conformal_full
predict.int_conformal_quantile <- function(object, new_data, ...) {
diff --git a/R/conformal_infer_split.R b/R/conformal_infer_split.R
index ddcfb4a..a8f38c3 100644
--- a/R/conformal_infer_split.R
+++ b/R/conformal_infer_split.R
@@ -94,6 +94,20 @@ print.int_conformal_split <- function(x, ...) {
invisible(x)
}
+#' @rdname required_pkgs.int_conformal_cv
+#' @keywords internal
+#' @export
+required_pkgs.int_conformal_split <- function(x, infra = TRUE, ...) {
+ model_pkgs <- required_pkgs(x$wflow, infra = infra)
+
+ if (infra) {
+ model_pkgs <- c(model_pkgs, "probably")
+ }
+
+ model_pkgs <- unique(model_pkgs)
+ model_pkgs
+}
+
#' @export
#' @rdname predict.int_conformal_full
predict.int_conformal_split <- function(object, new_data, level = 0.95, ...) {
diff --git a/man/int_conformal_full.Rd b/man/int_conformal_full.Rd
index 09faf20..e950fe7 100644
--- a/man/int_conformal_full.Rd
+++ b/man/int_conformal_full.Rd
@@ -68,6 +68,77 @@ the initial set of training set residuals are modeled using a Gamma generalized
linear model with a log link (see the reference by Aitkin below). For a new
sample, the absolute size of the residual is estimated and a multiple of
this value is computed as an initial guess of the search boundaries.
+
+\subsection{Speed}{
+
+The time it takes to compute the intervals depends on the training set
+size, search parameters (i.e., convergence criterion, number of
+iterations), the grid size, and the number of worker processes that are
+used. For the last item, the computations can be parallelized using the
+future and furrr packages.
+
+To use parallelism, the \code{\link[future:plan]{future::plan()}} function can
+be invoked to create a parallel backend. For example, let’s make an
+initial workflow:
+
+\if{html}{\out{}}\preformatted{library(tidymodels)
+library(probably)
+library(future)
+
+tidymodels_prefer()
+
+## Make a fitted workflow from some simulated data:
+set.seed(121)
+train_dat <- sim_regression(200)
+new_dat <- sim_regression( 5) |> select(-outcome)
+
+lm_fit <-
+ workflow() |>
+ add_model(linear_reg()) |>
+ add_formula(outcome ~ .) |>
+ fit(data = train_dat)
+
+# Create the object to be used to make prediction intervals
+lm_conform <- int_conformal_full(lm_fit, train_dat)
+}\if{html}{\out{
}}
+
+We’ll use a \code{"multisession"} parallel processing plan to compute the
+intervals for the five new samples in parallel:
+
+\if{html}{\out{}}\preformatted{plan("multisession")
+
+# This is run in parallel:
+predict(lm_conform, new_dat)
+}\if{html}{\out{
}}
+
+\if{html}{\out{}}\preformatted{## # A tibble: 5 x 2
+## .pred_lower .pred_upper
+##
+## 1 -17.9 59.6
+## 2 -33.7 51.1
+## 3 -30.6 48.2
+## 4 -17.3 59.6
+## 5 -23.3 55.2
+}\if{html}{\out{
}}
+
+Using simulations, there are slightly sub-linear speed-ups when using
+parallel processing to compute the row-wise intervals.
+
+In comparison with parametric intervals:
+
+\if{html}{\out{}}\preformatted{predict(lm_fit, new_dat, type = "pred_int")
+}\if{html}{\out{
}}
+
+\if{html}{\out{}}\preformatted{## # A tibble: 5 x 2
+## .pred_lower .pred_upper
+##
+## 1 -19.2 59.1
+## 2 -31.8 49.7
+## 3 -31.0 47.6
+## 4 -17.8 60.1
+## 5 -23.6 54.3
+}\if{html}{\out{
}}
+}
}
\references{
Jing Lei, Max G'Sell, Alessandro Rinaldo, Ryan J. Tibshirani and Larry
diff --git a/man/required_pkgs.int_conformal_cv.Rd b/man/required_pkgs.int_conformal_cv.Rd
new file mode 100644
index 0000000..e5d2d84
--- /dev/null
+++ b/man/required_pkgs.int_conformal_cv.Rd
@@ -0,0 +1,29 @@
+% Generated by roxygen2: do not edit by hand
+% Please edit documentation in R/conformal_infer.R, R/conformal_infer_cv.R,
+% R/conformal_infer_quantile.R, R/conformal_infer_split.R
+\name{required_pkgs.int_conformal_full}
+\alias{required_pkgs.int_conformal_full}
+\alias{required_pkgs.int_conformal_cv}
+\alias{required_pkgs.int_conformal_quantile}
+\alias{required_pkgs.int_conformal_split}
+\title{S3 methods to track which additional packages are needed for prediction
+intervals via conformal inference}
+\usage{
+\method{required_pkgs}{int_conformal_full}(x, infra = TRUE, ...)
+
+\method{required_pkgs}{int_conformal_cv}(x, infra = TRUE, ...)
+
+\method{required_pkgs}{int_conformal_quantile}(x, infra = TRUE, ...)
+
+\method{required_pkgs}{int_conformal_split}(x, infra = TRUE, ...)
+}
+\arguments{
+\item{x}{a conformal interval object}
+
+\item{...}{Other arguments passed to methods}
+}
+\description{
+S3 methods to track which additional packages are needed for prediction
+intervals via conformal inference
+}
+\keyword{internal}
diff --git a/tests/testthat/test-conformal-intervals-quantile.R b/tests/testthat/test-conformal-intervals-quantile.R
index c11102f..293c773 100644
--- a/tests/testthat/test-conformal-intervals-quantile.R
+++ b/tests/testthat/test-conformal-intervals-quantile.R
@@ -63,4 +63,13 @@ test_that("split conformal quantile intervals", {
nrow(new_int),
nrow(sim_new)
)
+
+ expect_identical(
+ required_pkgs(lm_int),
+ c(required_pkgs(lm_int$wflow), "probably")
+ )
+ expect_identical(
+ required_pkgs(lm_int, infra = FALSE),
+ required_pkgs(lm_int$wflow, infra = FALSE)
+ )
})
diff --git a/tests/testthat/test-conformal-intervals-split.R b/tests/testthat/test-conformal-intervals-split.R
index 280440f..6824e4e 100644
--- a/tests/testthat/test-conformal-intervals-split.R
+++ b/tests/testthat/test-conformal-intervals-split.R
@@ -64,4 +64,13 @@ test_that("split conformal intervals", {
nrow(new_int),
nrow(sim_new)
)
+
+ expect_identical(
+ required_pkgs(lm_int),
+ c(required_pkgs(lm_int$wflow), "probably")
+ )
+ expect_identical(
+ required_pkgs(lm_int, infra = FALSE),
+ required_pkgs(lm_int$wflow, infra = FALSE)
+ )
})
diff --git a/tests/testthat/test-conformal-intervals.R b/tests/testthat/test-conformal-intervals.R
index 467fada..5233d2e 100644
--- a/tests/testthat/test-conformal-intervals.R
+++ b/tests/testthat/test-conformal-intervals.R
@@ -229,6 +229,24 @@ test_that("conformal intervals", {
expect_true(
all(grid_bounds$.pred_lower < grid_bounds_90$.pred_lower)
)
+
+ expect_identical(
+ required_pkgs(smol_obj),
+ c(required_pkgs(smol_obj$wflow), "probably")
+ )
+ expect_identical(
+ required_pkgs(smol_obj, infra = FALSE),
+ required_pkgs(smol_obj$wflow, infra = FALSE)
+ )
+
+ expect_identical(
+ required_pkgs(grid_int),
+ c(unique(unlist(map(grid_int$models, required_pkgs))), "probably")
+ )
+ expect_identical(
+ required_pkgs(grid_int, infra = FALSE),
+ unique(unlist(map(grid_int$models, required_pkgs, infra = FALSE)))
+ )
})
test_that("conformal control", {