Package {portion}

portion: Extracting a Data Portion

Description

Extracts a portion of the data in vectors, matrices, data frames, time series, and lists of these objects. The size of the portion is given relative to the size of the object. The portion can consist of the first, the last, random, similar, or dissimilar elements, rows, or columns.

Author(s)

Maintainer: Lennart Oelschläger oelschlaeger.lennart@gmail.com

Authors:

• Lennart Oelschläger oelschlaeger.lennart@gmail.com

See Also

Useful links:

• https://github.com/loelschlaeger/portion

• Report bugs at https://github.com/loelschlaeger/portion/issues

Extract a data portion

Description

Extract a relative portion from vectors, matrices, data frames, time series, or lists of these objects.

Usage

portion(x, proportion, how = "random", centers = 2L, ...)

## Default S3 method:
portion(x, ...)

## S3 method for class 'numeric'
portion(x, proportion, how = "random", centers = 2L, ...)

## S3 method for class 'ts'
portion(x, proportion, how = "random", centers = 2L, byrow = TRUE, ...)

## S3 method for class 'character'
portion(x, proportion, how = "random", ...)

## S3 method for class 'logical'
portion(x, proportion, how = "random", centers = 2L, ...)

## S3 method for class 'matrix'
portion(
  x,
  proportion,
  how = "random",
  centers = 2L,
  byrow = TRUE,
  ignore = integer(),
  ...
)

## S3 method for class 'data.frame'
portion(
  x,
  proportion,
  how = "random",
  centers = 2L,
  byrow = TRUE,
  ignore = integer(),
  ...
)

## S3 method for class 'list'
portion(x, proportion, how = "random", centers = 2L, ...)

Arguments

Details

A portion of a vector consists of some of its elements, while a portion of a matrix or data frame consists of some of its rows (or columns if byrow = FALSE). The selected elements, rows, or columns keep their original order. For a list, each element is portioned separately, such that the selected indices can differ between the elements of the list.

Value

A portion of x, preserving attributes where possible. Vectors return selected elements, matrices and data frames return selected rows or columns, and lists return a list with each element portioned. The selected indices are stored in the "indices" attribute of each returned object.

Similar and dissimilar elements

Options how = "similar" and how = "dissimilar" are based on clustering the elements, rows, or columns via kmeans, which requires numeric or logical data without NA, NaN, or infinite values. Note that the clustering starts from randomly chosen centers, hence the result depends on the state of the random number generator (see set.seed).

Based on the clustering, the portion is selected as follows:

• For similar elements, the portion is filled cluster by cluster, starting with the largest cluster, such that it contains as few clusters as possible.

• For dissimilar elements, one element of each cluster is selected in turn until the portion is complete.

Examples

# can portion vectors, matrices, data.frames, and lists of such types
portion(
  list(
    1:10,
    matrix(LETTERS[1:12], nrow = 3, ncol = 4),
    data.frame(a = 1:6, b = -6:-1)
  ),
  proportion = 0.5,
  how = "first"
)

# can portion similar and dissimilar elements (based on kmeans clustering)
set.seed(1)
x <- c(1, 1, 2, 2)
portion(x, proportion = 0.5, how = "similar")
portion(x, proportion = 0.5, how = "dissimilar")

# ignore non-numeric columns when clustering data frame rows
x <- data.frame(value = c(1, 1, 5, 5), group = c("a", "a", "b", "b"))
portion(x, proportion = 0.5, how = "similar", ignore = 2)

# object attributes are preserved
x <- structure(1:10, "test_attribute" = "test")
x[1:5]
portion(x, proportion = 0.5, how = "first")

# can portion time series
portion(stats::ts(1:8, start = 2000), proportion = 0.5, how = "first")