R packages & functions
Two ways to compute the same agreement coefficients as AgreeStat360, from R: install the CRAN packages, or source individual script files directly.
irrCAC & irrICC
Both packages were developed by Dr. Gwet. irrCAC computes chance-corrected agreement coefficients (CAC) — Cohen's Kappa, Gwet's AC1/AC2, Krippendorff's alpha, and related measures. irrICC computes intraclass correlation coefficients (ICC) for quantitative ratings.
Both packages can be installed from CRAN like any standard R package, or directly from GitHub:
devtools::install_github("kgwet/irrCAC")
devtools::install_github("kgwet/irrICC")
Downloadable R script files
Prefer not to install a package? These script files each contain a set of R functions you can source directly, covering the same coefficients.
agree.coeff2.r
Computes agreement coefficients and their standard errors for 2 raters, with ratings organized as a contingency table (a square matrix of subject counts by rater and category). Works best when both raters rated every subject — otherwise subjects rated by only one rater may not be classified correctly.
agree.coeff3.dist.r
Computes agreement coefficients and their standard errors for multiple raters, with ratings organized as an n × q table — n subjects by q categories — showing how many raters assigned each subject to each category.
agree.coeff3.raw.r
Computes agreement coefficients and their standard errors for multiple raters, with ratings organized as an n × r table of raw (alphanumeric) ratings — n subjects, r raters, one row per subject. Requires weights.gen.r to also be loaded.
weights.gen.r
Generates the weight sets used for computing weighted agreement coefficients. Also required by paired.ttest.r below.
paired.ttest.r
Implements the paired t-test for testing whether two agreement coefficients differ significantly, based on the linearization method described in the 2016 paper on correlated agreement coefficients.