DeclareDesign is an R implementation of the MIDA framework, in which a research design has a Model of the world, an Inquiry, a Data strategy, and an Answer strategy. Each component is declared as a composable R function; designs are combined with + and diagnosed using Monte Carlo simulation to estimate properties such as power and bias.

What’s covered#

  • Model – declare_population() for simple and hierarchical datasets, declare_potential_outcomes()
  • Inquiry – declare_estimand() for causal, descriptive, and conditional estimands
  • Data strategy – declare_sampling() and declare_assignment() with stratification and clustering
  • Answer strategy – declare_estimator() for OLS, 2SLS, and difference-in-means
  • Design declaration – combining components with +
  • Design diagnosis – diagnose_design(), summary(), custom diagnosands