README

matrixCorr computes correlation and related association matrices from small to high-dimensional data using simple, consistent functions and sensible defaults. It includes shrinkage and robust options for noisy or p ≥ n settings, plus convenient print/plot methods. Performance-critical paths are implemented in C++ with BLAS/OpenMP and memory-aware symmetric updates. The API accepts base matrices and data frames and returns standard R objects via a consistent S3 interface.

Contributions from other researchers who want to add new correlation methods are very welcome. A central goal of matrixCorr is to keep efficient correlation and agreement estimation in one package with a common interface and consistent outputs, so methods can be extended, compared, and used without repeated translation across packages.

Supported measures include Pearson, Spearman, Kendall, distance correlation, partial correlation, and robust biweight mid-correlation; agreement tools cover Bland–Altman (two-method and repeated-measures) and Lin’s concordance correlation coefficient (including repeated-measures LMM/REML extensions).

Features

Installation

# Install from GitHub
# install.packages("devtools")
devtools::install_github("Prof-ThiagoOliveira/matrixCorr")

Example

Correlation matrices (Pearson, Spearman, Kendall)

library(matrixCorr)

set.seed(1)
X <- as.data.frame(matrix(rnorm(300 * 6), ncol = 6))
names(X) <- paste0("V", 1:6)

R_pear <- pearson_corr(X)
R_spr  <- spearman_rho(X)
R_ken  <- kendall_tau(X)

print(R_pear, digits = 2)
plot(R_spr)   # heatmap

Robust correlation (biweight mid-correlation)

set.seed(2)
Y <- X
# inject outliers
Y$V1[sample.int(nrow(Y), 8)] <- Y$V1[sample.int(nrow(Y), 8)] + 8

R_bicor <- biweight_mid_corr(Y)
print(R_bicor, digits = 2)

High-dimensional shrinkage correlation (\(p >> n\))

set.seed(3)
n <- 60; p <- 200
Xd <- matrix(rnorm(n * p), n, p)
colnames(Xd) <- paste0("G", seq_len(p))

R_shr <- schafer_corr(Xd)
print(R_shr, digits = 2, max_rows = 6, max_cols = 6)

Partial correlation matrix

R_part <- partial_correlation(X)
print(R_part, digits = 2)

Distance correlation matrix

R_dcor <- distance_corr(X)
print(R_dcor, digits = 2)

distance_corr() uses an unbiased estimator with a fast univariate (O(n n)) dispatch and an exact (O(n^2)) fallback for robustness.

Agreement analyses

Two-method Bland–Altman

set.seed(4)
x <- rnorm(120, 100, 10)
y <- x + 0.5 + rnorm(120, 0, 8)

ba <- bland_altman(x, y)
print(ba)
plot(ba)

Repeated-measures Bland–Altman (pairwise matrix)

set.seed(5)
S <- 20; Tm <- 6
subj  <- rep(seq_len(S), each = Tm)
time  <- rep(seq_len(Tm), times = S)

true  <- rnorm(S, 50, 6)[subj] + (time - mean(time)) * 0.4
mA    <- true + rnorm(length(true), 0, 2)
mB    <- true + 1.0 + rnorm(length(true), 0, 2.2)
mC    <- 0.95 * true + rnorm(length(true), 0, 2.5)

dat <- rbind(
  data.frame(y = mA, subject = subj, method = "A", time = time),
  data.frame(y = mB, subject = subj, method = "B", time = time),
  data.frame(y = mC, subject = subj, method = "C", time = time)
)
dat$method <- factor(dat$method, levels = c("A","B","C"))

ba_rep <- bland_altman_repeated(
  data = dat, response = "y", subject = "subject",
  method = "method", time = "time",
  include_slope = FALSE, use_ar1 = FALSE
)
summary(ba_rep)
# plot(ba_rep)  # faceted BA scatter by pair

Lin’s concordance correlation coefficient (repeated-measures LMM/REML)

set.seed(6)
S <- 30; Tm <- 8
id     <- factor(rep(seq_len(S), each = 2 * Tm))
method <- factor(rep(rep(c("A","B"), each = Tm), times = S))
time   <- rep(rep(seq_len(Tm), times = 2), times = S)

u  <- rnorm(S, 0, 0.8)[as.integer(id)]
g  <- rnorm(S * Tm, 0, 0.5)
g  <- g[ (as.integer(id) - 1L) * Tm + as.integer(time) ]
y  <- (method == "B") * 0.3 + u + g + rnorm(length(id), 0, 0.7)

dat_ccc <- data.frame(y, id, method, time)
fit_ccc <- ccc_lmm_reml(dat_ccc, response = "y", rind = "id",
                        method = "method", time = "time")
summary(fit_ccc)  # overall CCC, variance components, SEs/CI

Contributing

Issues and pull requests are welcome. Please see CONTRIBUTING.md for guidelines and cran-comments.md/DESCRIPTION for package metadata.

matrixCorr