In poissonconsulting/sims: Simulate Data from R or 'JAGS' Code

knitr::opts_chunk$set(
  collapse = TRUE,
  comment = "#>"
)

sims is an R package to generate datasets from JAGS or R code for use in simulation studies.

Simulating as nlists Objects

By default, sims_simulate() returns the simulated datasets in the form of an nlists object.

library(sims)
set.seed(10)
sims_simulate("a <- runif(1)", nsims = 2L)

Saving as .rds Files

If, however, save = TRUE then each nlist object is saved as an .rds file in path.

set.seed(10)
sims_simulate("a <- runif(1)", nsims = 2L, save = TRUE, path = tempdir(), exists = NA)
sims_data_files(tempdir())

Importing .rds Files

The datasets in the .rds files can be imported as an nlists object using sims_data().

sims_data(tempdir())

Adding .rds Files

The values including the .Random.seed (not printed) that were used to generate the datasets are saved in .sims_args.rds which can be imported using sims_info().

sims_info(tempdir())[1:5]

The fact that the arguments to sims_simulate() are saved to file allows additional .rds datasets to be generated using sims_add().

sims_add(tempdir(), nsims = 3L)
sims_data_files(tempdir())

Copying .rds Files

If the user wishes to duplicate the datasets then they can either regenerate them by specifying a different path but the same seed (using set.seed()). Alternatively, they can copy the existing .sims.rds and datasets files to a new directory using sims_copy()

sims_copy(path_from = tempdir(), path_to = paste0(tempdir(), "_copy"))

Checking .rds Files

A user can check that all the datasets specified in .sims.rds are present using sims_check().

sims_check(path = paste0(tempdir(), "_copy"))

file.remove(file.path(paste0(tempdir(), "_copy"), "data0000005.rds"))

sims_check(path = paste0(tempdir(), "_copy"))

unlink(paste0(tempdir(), "_copy"), recursive = TRUE, force = TRUE) # required to clean up detritus

Parallelization

Parallelization is implemented using the future package.

To use all available cores on the local machine simply execute the following code before calling sims_simulate().

library(future)
plan(multisession)

set.seed(10)
sims_simulate("a <- runif(1)", nsims = 2L)

Progress

Progress is reported using the progressr package as follows.

library(progressr)
with_progress(sims_simulate("a <- runif(1)", nsims = 1000L))

poissonconsulting/sims documentation built on June 18, 2022, 2:36 a.m.