The goal of saltr (pronounced salt-er, like the word jump in Spanish), is to help create readable simulations through the use of semantically meaningful functions. that allow researchers to use it in a flexible way without needing to commit fully to the package.
How is this library different?
A lot of the statistics libraries make heavy use of objects. This allows packages to handle things neatly and avoid errors that can be hard to debug on the user side. The downside is, however, that this closed systems are stiff black boxes, most of the time require a lot of custom code and have a steeper learning curve.
Saltr is a library that aims to give researchers, in the DCM field, the tools they need to jump start their simulations by making it easier to work with matrices and the gdina objects from the CDM package. Saltr has been designed following the tidyverse style guide so all functions are actions that are designed to work nicely with the pipe operator |>. Additionally, functions try and implement meaningful warning and error messages that help fix bugs during the coding process.
Core functions
Saltr is built around 5 function families identified by their suffix: create_*, build_*, generate_*, get_*, and change_*.
The create_* functions are mostly utility functions. They take some simple arguments and help create useful objects. For example, create_qmatrix makes it faster to create a starting Q-matrix that can be then modified.
The build_* functions always take as an input some matrix and return an object that is related. For example, we can create a test that is consistent with a Q-matrix by calling build_test_parameters. These functions can also be useful to build theoretical results like build_response_likelihood.
The generate_* functions are the only group that is not deterministic, hence the name. These functions are the bread and butter of a simulation study. We use these to create random samples of examinees (generate_examinees) and their responses (generate_responses).
So far all these families of functions are package agnostic, but in a simulation there will be a moment where we will want to fit a DCM model and extract some information. Some information is readily available, but other can be more cumbersome and require extra wrangling. The get_* functions simplify the extraction of information from the model.
Finally, the change_* family takes and object and returns a modified version of it. Useful for changes to objects that are relevant for a study like change_qmatrix and change_test_parameters that simplify studying, for example, misspecification of a model.
Get Started
You can install saltr from GitHub with:
devtools::install_github("beltranpantoja/saltR")And then jump to Creating a single simulation tutorial