Political scientists Graeme Blair, Jasper Cooper, Alexander Coppock, and Macartan Humphreys write:
DeclareDesign is a system for describing research designs in code and simulating them in order to understand their properties. Because DeclareDesign employs a consistent grammar of designs, you can focus on the intellectually challenging part – designing good research studies – without having to code up simulations from scratch. DeclareDesign is based on the Model-Inquiry-Data Strategy-Answer Strategy (MIDA) framework for describing designs and a declare-diagnose-redesign workflow for improving research designs before implementing them.
I haven’t used DeclareDesign myself, but I imagine it could be very helpful for researchers. Here’s an example with the 8 schools.
Plug from a user: I’ve found DeclareDesign very useful in thinking through research designs. Makes power calculations and much else very easy. Even if I’m not using the full suite of packages, fabricatr is great for generating artificial datasets, and estimatr is my go-to for robust standard errors in R.
Does the DeclareDesign ecosystem play nice with rstanarm?
Very interesting development!
There exists a designr package that allows design planning.
We also wrote a paper recently that shows code (R and brms/Stan) for planning one’s sample size for a Bayesian analysis: https://psyarxiv.com/u8yvc
In the vignette, i think it good to add a case for the 8 schhols analysis with regularizing priors otherwise it is a bit misleading for the beginner.
The MIDA framework has changed my perspective on empirical social science research. It’s cool to see the DD team get highlighted on here!