reval (version 1.0)

evalmany: Repeated evaluations

Description

Evaluate a function repeatedly across argument sets or permutations.

Usage

evalmany(fun, ..., method = c("ofat", "permute", "set"), sample = 0L,
  default.args = list(), collate = TRUE, collate.id = c("single",
  "multi"), collate.fun = identity, clusters = 1L)

Arguments

fun

The function to be evaluated.

...

Arguments to be varied when evaluating fun, where each argument in '...' is a (named) vector or list of values. Lists of multi-value objects (e.g. data.frames) should be named explicitly and may otherwise produce unexpected or

method

The sensitivity analysis method to be used. Can be either one-factor-at-a-time ("ofat") evaluation, evaluation of parameter sets ("set"), or (sampled) permutations of parameter sets ("permute"). When method = "ofat", the first element of e

sample

If method = "permute", the number of parameter permutations to evaluate (sampling without replacement). If sample < 1 (the default) then all possible permutations are evaluated.

default.args

The default values of arguments passed to fun.

collate

Whether to collate the results or not. If TRUE, output elements will be coerced into data.frames using as.data.frame. Otherwise, the raw outputs will be returned as a named list.

collate.id

If collate = TRUE, the method used to store the evaluation identifiers. If collate.id = "single", a single column is used. If collate.id = "multi", one column is created for each argument in '...'.

collate.fun

If collate = TRUE, an optional function for reshaping the output of each evaluation prior to coercing and collating the outputs.

clusters

Number of clusters to use for parallel processing. Default is 1 (serial computation).

Value

• If collate = TRUE, a data.frame. Otherwise, a named list.

Examples

set.seed(1459)
myfun = function(...){
  x = rnorm(...)
  data.frame(mean = mean(x), sd = sd(x))
}
evalmany(myfun, mean = c(5, 9), sd = c(2, 3), default.args = list(n = 1e6))
evalmany(myfun, mean = c(5, 9), sd = c(2, 3), default.args = list(n = 1e6),
  collate.id = "multi")
evalmany(myfun, mean = seq(20), sd = seq(1, 4, by = 0.1),
  default.args = list(n = 1e6), method = "permute", sample = 10)

# vector recycling
evalmany(myfun, mean = c(0, 3, 5), sd = c(1, 10),
  default.args = list(n = 1e6), method = "set", collate.id = "multi")

# Parallel processing
evalmany(myfun, mean = seq(0, 50, by = 10), sd = seq(1, 10, by = 1.5),
  default.args = list(n = 1e5), method = "permute", collate.id = "multi",
  clusters = 2)

# Complex objects and error handling
formulas = list(y ~ 1, y ~ x, y ~ I(x^2), y ~ x + z)
datasets = list(
  A = data.frame(x = seq(0, 99), y = seq(0, 99) + rnorm(100)),
  B = data.frame(x = seq(0, 99), y = seq(0, 99) + rnorm(100, mean = 5)),
  C = data.frame(x = seq(0, 99), y = seq(0, 99) + rlnorm(100, meanlog = 1))
)
# raw output
evalmany(lm, formula = formulas, data = datasets, method = "set",
  collate = FALSE)
# data extraction
evalmany(lm, formula = formulas, data = datasets, method = "permute",
  collate.id = "multi", collate.fun = function(x)
    data.frame(param = names(x$coefficients), value = x$coefficients,
    row.names=NULL))