lcc

lcc provides estimation routines for longitudinal concordance correlation (LCC), longitudinal Pearson correlation (LPC), and longitudinal accuracy (LA) via polynomial mixed-effects regression. With it you can quantify agreement profiles across methods, browse numeric and graphical summaries, and extract confidence intervals for each metric.

Key capabilities include:

• Unified interfaces for fitting agreement models and producing tidy summaries.
• Numerical and graphical diagnostics for fitted and sampled trajectories.
• Support for balanced or unbalanced designs, variance structures with time-dependent weights, and additional covariates in the fixed-effects predictor.

The project is maintained by Thiago de Paula Oliveira [cre, aut], Rafael de Andrade Moral [aut], John Hinde [aut], Silvio Sandoval Zocchi [ctb], and Clarice Garcia Borges Demétrio [ctb].

It has been available on CRAN since 2018 (https://CRAN.R-project.org/package=lcc). CRAN hosts the recommended stable release; see NEWS.md for the development change log.

Automated checks for the development branch run on Linux, macOS, and Windows via GitHub Actions, with code coverage reported through Codecov. To verify the test suite locally you can run:

devtools::test()
devtools::check()

This github page has its version under development. New functions will be added as experimental work and, once it is done and running correctly, we will synchronize the repositories and add it to the CRAN.

We worked hard to release a new stable version allowing users to analyze data sets, where the objective is studied the extent of the agreement profile among methods considering time as covariable.

lcc comprises a set of functions that help you build and summarise the fitted model, compute point estimates and bootstrap confidence intervals for LCC, LPC, and LA, and generate publication-ready plots. Some helper functions remain internal and are not intended for direct use.

Bootstrap enhancements

The latest development cycle introduced a richer bootstrap engine for agreement metrics:

• Confidence intervals now support normal, percentile, and bias-corrected and accelerated (BCa) estimators for LCC, LPC, and LA.
• Parallel resampling respects reproducible seeding via boot.seed and works across every supported bootstrap scheme.
• Degenerate-response safeguards ensure intervals collapse gracefully when the observed variability is insufficient for inference.
• Multiple resampling schemes are available through boot.scheme, covering subject-level, residual-based, semi-parametric, and fully parametric workflows.

fit <- lcc(
  data              = your_data,
  resp              = "response",
  subject           = "subject_id",
  method            = "assay",
  time              = "visit",
  ci                = TRUE,
  ci.method         = "bca",
  boot.scheme       = "np_case_resid_gr",
  nboot             = 1000,
  boot.seed         = 2025,
  numCore           = 4,
  keep.boot.models  = FALSE,
  components        = TRUE
)

summary(fit)
plot(fit)

Installation

Installed from CRAN:

install.packages("lcc")

Installed the development version from Github:

install.packages("devtools")
devtools::install_github("Prof-ThiagoOliveira/lcc")

If you use Windows, first install Rtools. If you run into issues, launch the installer with the Run as Administrator option. On macOS, install Xcode from the App Store.

lcc can also be installed by downloading the appropriate files directly at the CRAN web site and following the instructions given in the section 6.3 Installing Packages of the R Installation and Administration manual.

Longitudinal Concordance Correlation App

We hope you learn more about the LCC using the LCC App. The application helps illustrate how the model parameters influence the agreement profile over time. Have fun!

Tutorials

You can read lcc tutorials in our PeerJ article (https://doi.org/10.7717/peerj.9850) or by using the link below:

LCC paper.