ebci

This R package implements robust empirical Bayes confidence intervals from Armstrong, Kolesár, and Plagborg-Møller (2022). See the package manual for documentation of the package functions, and the package vignette for a description of the package and an example of its usage (available through vignette("ebci") once the package is installed). See ebci_matlab for a Matlab version of this package, and ebreg for a Stata version.

This software package is based upon work supported by the National Science Foundation under grant numbers SES-2049765 (Armstrong), SES-22049356 (Kolesár), and SES-1851665 (Plagborg-Møller), and by work supported by the Alfred P. Sloan Research Fellowship (Kolesár).

Installation

You can install the released version of ebci from CRAN with:

install.packages("ebci")

Alternatively, you can get the current development version from GitHub:

if (!requireNamespace("remotes")) {
  install.packages("remotes")
}
# remotes::install_github("kolesarm/ebci")
install.packages("ebci", repos = c('https://kolesarm.r-universe.dev', 'https://cloud.r-project.org'))

Example

Calculation of the critical values used to construct robust empirical Bayes confidence intervals:

library("ebci")
## Usual critical value
cva(m2=0, kappa=Inf, alpha=0.05)
## Larger critical value that takes bias into account. Only uses second moment
## constraint on normalized bias, m2=4.
cva(m2=4, kappa=Inf, alpha=0.05)
## Add a constraint that kurtosis equals 3. This tightens the critical value
cva(m2=4, kappa=3, alpha=0.05)

Estimates and robust EBCIs for neighborhood effects, as in the empirical application in Armstrong, Kolesár, and Plagborg-Møller (2020). Shrink fixed-effect estimates of the neighborhood effects, for children with parents at the 25th percentile of the income distribution (theta25) toward average outcome for permanent residents (stayers) at the 25th percentile of the income distribution. Use precision weights proportional to the inverse of the squared standard error of the fixed-effect estimates.

r <- ebci(theta25 ~ stayer25, data=cz, se=se25, weights=1/se25^2)