Conditional Standard Error of Measurement in Generalizability Theory • csemGT

The goal of csemGT is to estimate the conditional standard error of measurement (CSEM) within the Generalizability Theory framework for the univariate, single-facet, persons-by-items (p × i) crossed design, following Brennan (1998). It was created and is maintained by René Gempp, paralleling the Stata module gtcsem.

Unlike most other psychometric frameworks, Generalizability Theory distinguishes two types of conditional measurement error: the absolute CSEM, appropriate when decisions concern the absolute magnitude of a person’s score (for example, mastery classification against a fixed cutpoint), and the relative CSEM, appropriate when decisions concern comparisons among persons (for example, ranking or selection). csemGT estimates both.

Installation

You can install the development version of csemGT from GitHub with:

# Development installation
pak::pak("rgempp/csemGT")
# or
remotes::install_github("rgempp/csemGT", build_vignettes = TRUE)

Once on CRAN:

install.packages("csemGT")

Example

library(csemGT)
data(iowa_like)

fit <- csem_gt(iowa_like)
print(fit)
plot(fit, plot_type = "both", error_types = "absolute")

Citation

If you use csemGT in published work, please cite it as:

Gempp, R. (2026). csemGT: Conditional Standard Error of Measurement in Generalizability Theory. R package version 1.0.0. https://github.com/rgempp/csemGT

References

Brennan, R. L. (1998). Raw-score conditional standard errors of measurement in generalizability theory. Applied Psychological Measurement, 22(4), 307–331. https://doi.org/10.1177/014662169802200401