Graeme Blair, Jasper Cooper, Alex Coppock and I are developing a suite of R packages that provide tools to formally characterize research designs, assess their properties, and register them. The key inputs required are an abstract statement of the estimands and estimators as well as the data structure. The outputs include dummy datasets, dummy analyses, and reports on properties of the design such as the power, the expected bias, and the coverage properties of estimators. These in turn can be used to select a design and generate a preregistration design document.
You might be surprised using it to see how simple designs may have fundamental flaws.
Lots more on the DeclareDesign here
For components, see:
- randomizr (for R and for Stata): Easy-to-use tools for common forms of random assignment and sampling.
- fabricatr: Imagine your data before you collect it.
- estimatr: Fast estimators for design-based inference.
- DeclareDesign: Declare and diagnose research designs.
- DesignLibrary: Templates for quick design declaration