fastconley: Fast Conley Standard Errors for 'lfe' and 'fixest' Models

Conley (1999) <doi:10.1016/S0304-4076(98)00084-0> spatial heteroscedasticity and autocorrelation consistent (HAC) standard errors for fixed effects panel and cross-sectional models estimated with felm() from the 'lfe' package (ordinary least squares and instrumental variables) or with feols(), feglm(), and fepois() from the 'fixest' package. Instrumental-variable support is limited to ordinary two-stage least squares. Generalized linear model fits use the M-estimation sandwich built from the stored scores and inverse Hessian. The spatial path uses score accumulation, a three-dimensional cell-grid neighbour search, and compressed sparse row neighbour lists instead of dense distance matrices, yielding large speedups over the original 'conley' package <https://github.com/rbluhm/conley> on big cross-sections and high-dimensional regressions.

Version: 0.11.1
Depends: R (≥ 4.0)
Imports: data.table, Rcpp, stats
LinkingTo: Rcpp, RcppArmadillo
Suggests: fixest, knitr, lfe, rmarkdown, testthat (≥ 3.0.0)
Published: 2026-09-26
DOI: 10.32614/CRAN.package.fastconley (may not be active yet)
Author: Richard Bluhm [aut, cre, cph]
Maintainer: Richard Bluhm <richard.bluhm at gmail.com>
BugReports: https://github.com/rbluhm/fastconley/issues
License: MIT + file LICENSE
URL: https://github.com/rbluhm/fastconley, https://rbluhm.github.io/fastconley/
NeedsCompilation: yes
Materials: README, NEWS
CRAN checks: fastconley results

Documentation:

Reference manual: fastconley.html , fastconley.pdf
Vignettes: Conley standard errors with fastconley (source, R code)

Downloads:

Package source: fastconley_0.11.1.tar.gz
Windows binaries: r-devel: not available, r-release: not available, r-oldrel: not available
macOS binaries: r-release (arm64): fastconley_0.11.1.tgz, r-oldrel (arm64): fastconley_0.11.1.tgz, r-release (x86_64): fastconley_0.11.1.tgz, r-oldrel (x86_64): fastconley_0.11.1.tgz

Linking:

Please use the canonical form https://CRAN.R-project.org/package=fastconley to link to this page.