| Type: | Package |
| Title: | Design-Based Causal Inference Method for Incomplete Block Designs |
| Version: | 0.0.1 |
| Description: | This R package implements methods for estimation and inference under Incomplete Block Designs and Balanced Incomplete Block Designs within a design-based finite-population framework. Based on 'Koo and Pashley' (2024) <doi:10.48550/arXiv.2405.19312>, it includes block-level estimators and extends to unit-level effects using 'Horvitz-Thompson' and 'Hájek' estimators. The package also provides asymptotic confidence intervals to support valid statistical inference. |
| License: | MIT + file LICENSE |
| Encoding: | UTF-8 |
| RoxygenNote: | 7.3.2 |
| Imports: | crossdes, dplyr, tidyr |
| URL: | https://github.com/taehyeonkoo/IBDInfer |
| NeedsCompilation: | no |
| Packaged: | 2025-04-16 23:02:26 UTC; taehyeon |
| Author: | Taehyeon Koo [aut, cre], Nicole Pashley [ctb] |
| Maintainer: | Taehyeon Koo <tk587@stat.rutgers.edu> |
| Repository: | CRAN |
| Date/Publication: | 2025-04-17 00:10:07 UTC |
Design-based Inference for Incomplete Block Designs
Description
Conduct the design-based inference for incomplete block designs.
Usage
IBDInfer(y, b, z, g, w = c("Unit", "Block"), alpha = 0.05, data = NULL)
Arguments
y |
Observed outcomes. |
b |
Block identifier (ID). |
z |
Assigned treatments. |
g |
A contrast vector, must sum to zero. |
w |
A weight vector, must sum to one and contain non-negative values. |
alpha |
Confidence level, default set to 0.05. |
data |
A data frame; if provided, y, b, and z should be column names in the data frame. |
Value
IBDInfer returns an object of class "IBD", which is a list containing the following components: :
tau.ht |
The Horvitz-Thompson estimator of tau. |
tau.haj |
The Hajek estimator of tau. |
var_tau_ht_bb |
Variance estimator for the Horvitz-Thompson estimator with between-block bias. |
var_tau_ht_wb |
Variance estimator for the Horvitz-Thompson estimator with within-block bias. |
var_tau_haj_bb |
Variance estimator for the Hajek estimator with between-block bias. |
var_tau_haj_wb |
Variance estimator for the Hajek estimator with within-block bias. |
CI_ht_bb |
Confidence interval with the Horvitz-Thompson estimator and variance estimator with between-block bias. |
CI_ht_wb |
Confidence interval with the Horvitz-Thompson estimator and variance estimator with within-block bias. |
CI_haj_bb |
Confidence interval with the Hajek estimator and variance estimator with between-block bias. |
CI_haj_wb |
Confidence interval with the Hajek estimator and variance estimator with within-block bias. |
yht |
The Horvitz-Thompson estimator for each treatment. |
yhaj |
The Hajek estimator for each treatment. |
Sht_bb |
Covariance estimator for the Horvitz-Thompson estimator for each treatment with between-block bias. |
Sht_wb |
Covariance estimator for the Horvitz-Thompson estimator for each treatment with within-block bias. |
Shaj_bb |
Covariance estimator for the Hajek estimator for each treatment with between-block bias. |
Shaj_wb |
Covariance estimator for the Hajek estimator for each treatment with within-block bias. |
alpha |
Confidence level |
References
Koo, T., Pashley, N.E. (2024), Design-based Causal Inference for Incomplete Block Designs, arXiv preprint arXiv:2405.19312.
Examples
K <- 6
n.trt <- 3
t <- 2
n.vec <- rep(4, K)
df <- IBDgen(K = K, n.trt = n.trt, t = t, n.vec = n.vec)$blk_assign
df$y <- rnorm(nrow(df), 0, 1)
IBDInfer <- IBDInfer(y = y, b = blk_id, z = assign, g = c(1, -1, 0), w = "Block", data = df)
Generating Incomplete Block Designs
Description
Generate incomplete block designs.
Usage
IBDgen(K, n.trt, t, n.vec = NULL, L = NULL, l = NULL, W = NULL, balanced = T)
Arguments
K |
The number of blocks. |
n.trt |
The number of whole treatments. |
t |
The number of treatments to be assigned to each block. |
n.vec |
The vector of block sizes. |
L |
The vector of the number of blocks having each treatment. |
l |
The matrix of the number of blocks having each pair of treatments. |
W |
The set of treatment subsets used in the design. |
balanced |
Whether the design is balanced or not. If |
Value
A list containing the following components:
W |
The set of treatment subsets used in the design. |
W.uniq |
The unique set of treatment subsets used in the design with proportion in |
Rk |
The block assignment matrix. |
blk_assign |
The block assignment data frame. |
References
Sailer, M. O., & Bornkamp, M. B. (2022). Package ‘crossdes’: Construction of Crossover Designs.
Examples
K <- 6
n.trt <- 3
t <- 2
n.vec <- rep(4, K)
IBDgen(K = K, n.trt = n.trt, t = t, n.vec = n.vec)
Global Variables for IBDInfer
Description
This section declares global variables used in the IBDInfer package to prevent R CMD check warnings.
Summary of IBD
Description
Summary function for IBDInfer
Usage
## S3 method for class 'IBD'
summary(object, ...)
Value
No return value, called for summary.