RMarkdown templates for research design documentation.
# Install release version from CRAN install.packages("researchr") # Install development version from GitHub devtools::install_github("andrewcstewart/researchr")
The primary purpose of researchr
is to aid in the design of research projects, particularly data science projects, by offering a lightweight composable frameork of document templates and functions to track and manage a research design throughout the course of its history.
The emergence of the data science Notebook and its ensuing ecosystem of tutorials tends to create the impression that data science can always be captured in a concise, linear sequence of steps within a single session. It makes sense for those tutorials to consist of single notebooks rather than require readers to reference a series of separate notebooks, but in practice real life is much more messy. Research projects typically span long lengths of time, attempt several different approaches and combinations of approaches, and iteratively build off of each previous version of work. A more accurate picture might be a network of notebooks ; and that’s basically what researchr
is.
There’s a couple key features that researchr
offers to improve the scientist’s quality of life:
There are plenty of tools out there to help data science projects including workflow managers, data versioning tools, metric collectors, reproducibility automation, etc. researchr
doesn’t attempt to do any of those things, but it should compliment any choice of those tools nicely.
At its core, the package consists of two primary components:
The research design itself consists simply of an RStudio project and some yaml
encoded metadata files, as well as RMarkdown files that document the research.
use_researchr
new_research
add_data
add_method
add_instrument
add_experiment
add_prototype
_design_inventory
- traverses the directory tree to generate an inventory of current design state.here
for “M-1”)