Computes weighted rank-dependent concentration and inequality
indices with explicit treatment of tied ranks. Implements the relative,
generalized, modified, Wagstaff-normalized and Erreygers-normalized
estimators described by O'Donnell et al. (2016)
This package provides weighted fractional-rank concentration indices (relative,
generalized, modified, Wagstaff and Erreygers) and supports replicate-weight
variance estimation through the survey package. It deliberately does not
claim numerical parity with every feature of Stata's conindex command.
install.packages("testthat")
install.packages("survey")
install.packages("devtools")
devtools::install("path/to/polecon")
library(polecon)
d <- data.frame(income = 1:4, fvi = c(0, 0, 1, 1), w = 1)
conindex(d, "fvi", "income", weight = "w", bounded = TRUE,
limits = c(0, 1), erreygers = TRUE)
The user must select a valid replicate construction for their sample design. For example, for a stratified clustered design with valid multi-PSU strata:
library(survey)
set.seed(42)
d <- data.frame(
strata = rep(1:3, each = 8),
psu = rep(1:12, each = 2),
income = seq_len(24),
fvi = seq(0.15, 0.85, length.out = 24),
wt = rep(c(1, 2), 12)
)
s <- svydesign(id = ~psu, strata = ~strata, weights = ~wt,
data = d, nest = TRUE)
sr <- as.svrepdesign(s, type = "bootstrap", replicates = 100, mse = TRUE)
out <- conindex(d, "fvi", "income", bounded = TRUE,
limits = c(0, 1), erreygers = TRUE, svy = sr)
print(out)
confint(out)
This is an initial, restricted, auditable foundation. Complex index families, Stata's extended/symmetric indices, regression-based robust/clustered standard errors and group comparisons from the originally supplied prototype have not been carried into this candidate, pending separate parity and reference tests. The code does not recalculate a separately estimated FVI or its category elasticities in each survey replicate: that must be implemented in the researcher's own resampling pipeline if such uncertainty is of interest.
Confirm the maintainer email in DESCRIPTION, verify the name
against CRAN at submission time, complete independent Stata parity tests,
run the testthat suite, and run R CMD build, then R CMD check --as-cran
on the resulting source tarball using a current version of R. This environment
did not have an R executable, so the package has not passed R checks here.
Concept and initial implementation follow O'Donnell, O'Neill, Van Ourti and
Walsh (2016), The Stata Journal, https://doi.org/10.1177/1536867X1601600112.
Their Stata code is distributed under GPL-3; see inst/COPYRIGHTS.
The project-specific example for food vulnerability indices is available in
inst/FVI_USAGE.md.