make_dilated_out_hh: Create potential dilated outcomes under a two-stage...

In szonszein/interference: Design-Based Estimation of Spillover Effects

Description

Create potential dilated outcomes for simulation under two-stage hierarchical treatment assignment, assuming partial and stratified interference.

Usage

make_dilated_out_hh(
  N,
  out = function(x) abs(rnorm(x)),
  multipliers = NULL,
  seed = NULL
)

Arguments

N	number of units.
out	function which returns a vector length N of outcome values for treatment condition Indirect Phi which is the baseline (or control) condition. Default function draws from an absolute standard normal distribution.
multipliers	numeric vector with dilated effects multipliers. Must be of length 3. Default is c(2,1.5,1.25).
seed	random number for result replicability.

Details

make_dilated_out_hh produces potential dilated outcomes for simulation according to a two-stage hierarchical design where groups are first randomly assigned to a high or a low level of treatment saturation (psi, phi), and then units within groups are randomly assigned to treatment with probability equal to their group saturation rate. Following Hudgens and Halloran (2008), there are four potential outcomes which correspond to the four potential treatment conditions: Direct + Indirect Psi–a unit is directly treated and its group is assigned treatment saturation psi, Direct + Indirect Phi–a unit is directly treated and its group is assigned treatment saturation phi, Indirect Psi–a unit is not directly treated and its group is assigned treatment saturation psi, Indirect Phi–a unit is not directly treated and its group is assigned treatment saturation phi. This function assumes stratified interference (i.e. potential outcomes of a unit are affected by its own treatment assignment and only the treated proportion of its group; the precise set of treated group members does not matter).

Value

An K * N named numeric matrix, where K corresponds to the number of exposure conditions and N number of units.

References

Hudgens, M.G. & Halloran M.E. (2008). Toward causal inference with interference. Journal of the American Statistical Association, 103(482), 832–842.

Aronow, P.M. et al. (2020). Spillover effects in experimental data. arXiv preprint, arXiv:2001.05444.

Examples

make_dilated_out_hh(N = 10, seed = 357)

multipliers <- c(4, 3, 2)
make_dilated_out_hh(N = 10, multipliers = multipliers, seed = 357)

szonszein/interference documentation built on Jan. 10, 2022, 6:35 p.m.