In tidymodels/rsample: General Resampling Infrastructure

knitr::opts_chunk$set(
  message = FALSE,
  digits = 3,
  collapse = TRUE,
  comment = "#>"
  )
options(digits = 3)

Terminology

We define a resample as the result of a two-way split of a data set. For example, when bootstrapping, one part of the resample is a sample with replacement of the original data. The other part of the split contains the instances that were not contained in the bootstrap sample. Cross-validation is another type of resampling.

rset Objects Contain Many Resamples

The main class in the package (rset) is for a set or collection of resamples. In 10-fold cross-validation, the set would consist of the 10 different resamples of the original data.

Like modelr, the resamples are stored in data-frame-like tibble object. As a simple example, here is a small set of bootstraps of the mtcars data:

library(rsample)
set.seed(8584)
bt_resamples <- bootstraps(mtcars, times = 3)
bt_resamples

Individual Resamples are rsplit Objects

The resamples are stored in the splits column in an object that has class rsplit.

In this package we use the following terminology for the two partitions that comprise a resample:

• The analysis data are those that we selected in the resample. For a bootstrap, this is the sample with replacement. For 10-fold cross-validation, this is the 90% of the data. These data are often used to fit a model or calculate a statistic in traditional bootstrapping.
• The assessment data are usually the section of the original data not covered by the analysis set. Again, in 10-fold CV, this is the 10% held out. These data are often used to evaluate the performance of a model that was fit to the analysis data.

(Aside: While some might use the term "training" and "testing" for these data sets, we avoid them since those labels often conflict with the data that result from an initial partition of the data that is typically done before resampling. The training/test split can be conducted using the initial_split() function in this package.)

Let's look at one of the rsplit objects

first_resample <- bt_resamples$splits[[1]]
first_resample

This indicates that there were r dim(bt_resamples$splits[[1]])["analysis"] data points in the analysis set, r dim(bt_resamples$splits[[1]])["assessment"] instances were in the assessment set, and that the original data contained r dim(bt_resamples$splits[[1]])["n"] data points. These results can also be determined using the dim function on an rsplit object.

To obtain either of these data sets from an rsplit, the as.data.frame() function can be used. By default, the analysis set is returned but the data option can be used to return the assessment data:

head(as.data.frame(first_resample))
as.data.frame(first_resample, data = "assessment")

Alternatively, you can use the shortcuts analysis(first_resample) and assessment(first_resample).

tidymodels/rsample documentation built on Sept. 29, 2024, 10:48 p.m.