Difference Values of CCC Between Implementations

[nathanhaigh]

Can you explain why the CCC calculated by condordancer is different to that calculated by two other packages?

library("concordancer")
library("DescTools")
library("epiR")

# Create x and y vectors
x <- rnorm(100, sd = 1)
y <- rnorm(length(x), sd = 0.2)

# Calculate CCC using 3 different packages
c <- concordancer::ccc(x, y)
d <- DescTools::CCC(x, y)
e <- epiR::epi.ccc(x, y)

# Look at their CCC
c
[1] -0.01593504
d$rho.c["est"]
          est
1 -0.01593447
e$rho.c["est"]
          est
1 -0.01593447

[nathanhaigh]

In case you want to provide microbenchmark results:

library(microbenchmark)
library(ggplot2)

library(concordancer)
library(DescTools)
library(epiR)

# Create x and y vectors
x <- rnorm(100, sd = 1)
y <- rnorm(length(x), sd = 0.2)

# Perform microbenchmarking
ccc_mbm <- microbenchmark(
  concordancer = concordancer::ccc(x, y),
  DescTools = DescTools::CCC(x, y)$rho.c["est"],
  epiR = epiR::epi.ccc(x, y)$rho.c["est"],
  
  check = NULL
)

# Plot microbenchmarking result
autoplot(ccc_mbm)

However, the comparison might not be "fair" since both DescTools and epiR also calculate and return other information such as confidence intervals.