rvar-slice: Random variable slicing

In posterior: Tools for Working with Posterior Distributions

Random variable slicing

Description

Operations for slicing rvars and replacing parts of rvars.

Usage

## S3 method for class 'rvar'
x[[i, ...]]

## S3 replacement method for class 'rvar'
x[[i, ...]] <- value

## S3 method for class 'rvar'
x[..., drop = FALSE]

## S3 replacement method for class 'rvar'
x[i, ...] <- value

Arguments

x	an rvar.
i, ...	indices; see Details.
value	(rvar or coercable to rvar) Value to insert into x at the location determined by the indices.
drop	(logical) Should singular dimensions be dropped when slicing array rvars? Unlike base array slicing operations, defaults to FALSE.

Details

The rvar slicing operators ([ and [[) attempt to implement the same semantics as the base array slicing operators. There are some exceptions; most notably, rvar slicing defaults to drop = FALSE instead of drop = TRUE.

Extracting or replacing single elements with [[

The [[ operator extracts (or replaces) single elements. It always returns (or replaces) a scalar (length-1) rvar.

The x[[i,...]] operator can be used as follows:

• ⁠x[[<numeric>]]⁠ for scalar numeric i: gives the ith element of x. If x is multidimensional (i.e. length(dim(x)) > 1), extra dimensions are ignored when indexing. For example, if x is a 6 \times 2 rvar array, the 7th element, x[[7]], will be the first element of the second column, x[1,2].

• ⁠x[[<numeric rvar>]]⁠ for scalar numeric rvar i: a generalization of indexing when i is a scalar numeric. Within each draw of x, selects the element corresponding to the value of i within that same draw.

• ⁠x[[<character>]]⁠ for scalar character i: gives the element of x with name equal to i. Unlike with base arrays, does not work with multidimensional rvars.

• x[[i_1,i_2,...,i_n]] for scalar numeric or character i_1, i_2, etc. Must provide exactly the same number of indices as dimensions in x. Selects the element at the corresponding position in the rvar by number and/or dimname (as a string).

Extracting or replacing multiple elements with [

The [ operator extracts (or replaces) multiple elements. It always returns (or replaces) a possibly-multidimensional rvar.

The x[i,...] operator can be used as follows:

• ⁠x[<logical>]⁠ for vector logical i: i is recycled to the same length as x, ignoring multiple dimensions in x, then an rvar vector is returned containing the elements in x where i is TRUE.

• ⁠x[<logical rvar>]⁠ for scalar logical rvar i: returns an rvar the same shape as x containing only those draws where i is TRUE.

• ⁠x[<numeric>]⁠ for vector numeric i: an rvar vector is returned containing the ith elements of x, ignoring dimensions.

• ⁠x[<matrix>]⁠ for numeric matrix i, where ncol(i) == length(dim(x)): each row of i should give the multidimensional index for a single element in x. The result is an rvar vector of length nrow(i) containing elements of x selected by each row of i.

• x[i_1,i_2,...,i_n] for vector numeric, character, or logical i_1, i_2, etc. Returns a slice of x containing all elements from the dimensions specified in i_1, i_2, etc. If an argument is left empty, all elements from that dimension are included. Unlike base arrays, trailing dimensions can be omitted entirely and will still be selected; for example, if x has three dimensions, both x[1,,] and x[1,] can be used to create a slice that includes all elements from the last two dimensions. Unlike base arrays, [ defaults to drop = FALSE, so results retain the same number of dimensions as x.

Examples

x <- rvar(array(1:24, dim = c(4,2,3)))
dimnames(x) <- list(c("a","b"), c("d","e","f"))
x

## Slicing single elements
# x[[<numeric>]]
x[[2]]

# x[[<numeric rvar>]]
# notice the draws of x[1:4]...
draws_of(x[1:4])
x[[rvar(c(1,3,4,4))]]
# ... x[[rvar(c(1,3,4,4))]] creates a mixures of those draws
draws_of(x[[rvar(c(1,3,4,4))]])

# x[[i_1,i_2,...]]
x[[2,"e"]]


## Slicing multiple elements
# x[<logical>]
x[c(TRUE,TRUE,FALSE)]

# x[<logical rvar>]
# select every other draw
x[rvar(c(TRUE,FALSE,TRUE,FALSE))]

# x[<numeric>]
x[1:3]

# x[<matrix>]
x[rbind(
  c(1,2),
  c(1,3),
  c(2,2)
)]

# x[i_1,i_2,...,i_n]
x[1,]
x[1,2:3]
x[,2:3]

posterior documentation built on Nov. 2, 2023, 5:56 p.m.