Getting Started with the Simulator

Starting from a template

After installing the package, open R and type.

library(simulator)
dir <- "./sims"
create(dir)

## New simulation template created!  Go to ./sims/main.R to get started.

Choose dir to be the path of a directory (that does not yet exist) where you want your simulation code and files to be stored. In practice, "./sims" would be a standard choice, where "." refers to a directory containing files relevant to your current project.

The create command generates a skeleton of a simulation.¹ A look at the newly created directory shows that several files have been created.

setwd(dir)
list.files()

## [1] "eval_functions.R"   "main.R"             "method_functions.R"
## [4] "model_functions.R"  "writeup.Rmd"

This is the template of a basic simulation.

• In model_functions.R, write code that defines the models under which you wish to simulate.
  
• In method_functions.R, add code for methods that you wish to compare in your simulation (note that by using source and library, you can keep method_functions.R short and to the point, focusing on calling new_method rather than putting the actual heart of algorithms in that file).
  
• In eval_functions.R, use new_metric to define the ways in which your methods will be evaluated.
• The file main.R contains the main entry point to the simulation. Running the code in this file determines which models/methods/metrics are computed, etc.
  
• Finally, the file writeup.Rmd shows how all results can be presented in as a report. This document pulls all code from the .R files mentioned above, so that as main.R and other files develop, the report will remain up to date. To create an html file report, run the following command in R (which requires installing the package rmarkdown).

rmarkdown::render("writeup.Rmd", "html_document")

Or if one is using RStudio, one can simply press the Knit HTML button.

Typical workflow

On a typical project, one starts by defining a model in model_functions.R, one or two methods in method_functions.R, and a few metrics in eval_functions.R, and then one runs the code in main.R. After looking at some of the results, one might add an additional model or method or metric. One then returns to main.R, adds some additional lines specifying that the additional components should be run as well and looks at some more results.

The simplest way to look at results is by using the plot functions plot_eval, plot_evals and plot_evals_by. In situations where you wish to investigate results more deeply than just looking at aggregated plots, one can use the functions model, draws, output, and evals to get at all objects generated through the course of the simulation.

Next steps

The best way to get a sense of how to use the simulator is to look at examples. There are several vignettes that demonstrate how the simulator can be used to conduct simulations for some of the most famous statistical methods.

1. Lasso vignette: Explains basics, including the magrittr pipe and making plots and tables. Also demonstrates some more advanced features such as writing method extensions (such as refitting the result of the lasso or performing cross-validation).
2. James-Stein vignette: Shows how to step into specific parts of the simulation for troubleshooting your code.
3. Elastic net vignette: Shows how we can work with a sequence of methods that are identical except for a parameter that varies
4. Benjamini-Hochberg vignette: Shows how we can load a preexisting simulation and add more random draws without having to rerun anything. It also shows how one can have multiple simulation objects that point to overlapping sets of results.