generate_blocked_data: Make simulated dataset from a list of block sizes

In lmiratrix/blkvar: ATE and Treatment Variation Estimation for Blocked and Multisite RCTs

Make simulated dataset from a list of block sizes

Description

This method is the one that generates the simulation data used in Pashley & Miratrix.

Generate data, form_blocks_from_continuous, and randomize within block and generate observed potential outcomes

Usage

generate_blocked_data(
  n_k,
  sigma_alpha = 1,
  sigma_beta = 0,
  beta = 5,
  sigma_0 = 1,
  sigma_1 = 1,
  corr = 0.5,
  exact = FALSE
)

generate_blocked_data_obs(n_k = c(2, 3, 4, 8), p = 0.5, ...)

Arguments

n_k	List of block sizes
sigma_alpha	Standard deviation of the block mean Y0s.
sigma_beta	Standard deviation of the block mean treatment effects (Y1-Y0)s.
beta	Block Average ATE.
sigma_0	Standard deviation of residual Y0 added to block means (can be vector for individual variances per block).
sigma_1	As 'sigma_0' but for Y1s.
corr	Correlation of Y0, Y1 within a block (can be vector of length K for different blocks).
exact	Passed to mvrnorm to control how block means are generated.
p	Proportion of units treated (as close as possible given block sizes). This can be a vector with a probability for each block.
...	Parameters to be passed to generate_blocked_data()

Details

The block means are sampled from a multivariate normal distribution. This can be controlled so the variances are exact using the 'exact' flag.

Value

Dataframe with block indicators, Y0, and Y1.

Dataframe with original potential outcomes and observed outcome based on random assigment.

lmiratrix/blkvar documentation built on Nov. 18, 2024, 1:27 p.m.