DeclareDesign: Declare and Diagnose Research Designs
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.
To install the latest stable release of DeclareDesign, please ensure that you are running version 3.3 or later of R and run the following code:
To install the latest development release of all of the packages, please ensure that you are running version 3.4 or later of R and run the following code:
install.packages("DeclareDesign", dependencies = TRUE, repos = c("http://R.declaredesign.org", "https://cloud.r-project.org"))
Designs are declared by adding together design elements. Here’s a minimal example that describes a 100 unit randomized controlled trial with a binary outcome. Half the units are assigned to treatment and the remainder to control. The true value of the average treatment effect is 0.05 and it will be estimated with the difference-in-means estimator. The diagnosis shows that the study is unbiased but underpowered.
library(DeclareDesign) design <- declare_population(N = 100) + declare_potential_outcomes(Y ~ rbinom(n = N, size = 1, prob = 0.5 + 0.05 * Z)) + declare_estimand(ATE = 0.05) + declare_assignment(m = 50) + declare_estimator(Y ~ Z) diagnosis <- diagnose_design(design, diagnosands = declare_diagnosands(select = c("power", "bias"))) diagnosis
## ## Research design diagnosis based on 500 simulations. Diagnosand estimates with bootstrapped standard errors in parentheses (100 replicates). ## ## Design Label Estimand Label Estimator Label Term N Sims Power Bias ## design ATE estimator Z 500 0.08 -0.00 ## (0.01) (0.00)
The core DeclareDesign package relies on three companion packages, each of which is useful in its own right.
- randomizr: Easy to use tools for common forms of random assignment and sampling.
- fabricatr: Imagine your data before you collect it.
- estimatr: Fast estimators for social scientists.
To get started, have a look at this vignette on the idea behind DeclareDesign, which covers the main functionality of the software.
You can also browse a library of already declared designs. The library includes canonical designs that you can download, modify, and deploy.
A fuller description of the philosophy underlying the software is described in this working paper.
Each of these
declare_*() functions returns a function.
declare_population()(describes dimensions and distributions over the variables in the population)
declare_potential_outcomes()(takes population or sample and adds potential outcomes produced by interventions)
declare_sampling()(takes a population and selects a sample)
declare_assignment()(takes a population or sample and adds treatment assignments)
declare_estimand()(takes potential outcomes and calculates a quantity of interest)
declare_estimator()(takes data produced by sampling and assignment and returns estimates)
To declare a design, connect the components of your design with the + operator.
Once you have declared your design, there are four core post-design-declaration commands used to modify or diagnose your design:
diagnose_design()(takes a design, returns simulations and diagnosis)
draw_data()(takes a design and returns a single draw of the data)
get_estimates()(takes a design a returns a single simulation of estimates)
get_estimands()(takes a design a returns a single simulation of estimands)
A few other features:
- A designer is a function that takes parameters (e.g.,
N) and returns a design.
expand_design()is a function of a designer and parameters that returns a design.
- You can change the features of the design to be diagnosed with
declare_reveal()implements a general switching equation, which allows you to reveal outcomes from potential outcomes and a treatment assignment.
- You can provide custom functions to any
declare_*step, as described in the custom functions vignette.