pairedmcnemar

Pairwise McNemar tests vs a control for paired binary outcomes, plus Cohen's g with confidence intervals.

Features

• Compares Control to 2+ treatments using contingencytables McNemar tests:
  • Asymptotic (default)
  • Asymptotic with continuity correction
  • Mid-P (reports Z = NA)
• Adjusts p-values across comparisons (Holm default; BH / Hochberg optional)
• Computes Cohen's g (effectsize) + CI
• Adds helpful metadata columns:
  • method_label, p_adjust_label, ci_label
  • settings combined label (short/full)
  • nd discordant pairs
• S3 summary() + print() methods with flexible display.

Install (development)

# install.packages("remotes")
remotes::install_github("PHUoL/pairedmcnemar")

Quick start

library(pairedmcnemar)

set.seed(1)
n <- 40
id <- 1:n
control <- rbinom(n, 1, 0.4)
A <- control
B <- ifelse(runif(n) < 0.25, 1 - control, control)
C <- rbinom(n, 1, 0.6)

example_dat <- data.frame(
  id = rep(id, 4),
  condition = rep(c("Control", "A", "B", "C"), each = n),
  outcome = c(control, A, B, C)
)

fit <- mcnemar_vs_control(
  example_dat, id, condition, outcome,
  mcnemar_method = "midp",
  p_adjust = "BH",
  ci = 0.90,
  settings_style = "short",
  settings = "yes"
)

# print uses the stored default display
fit

# override display
summary(fit, settings = "none")
summary(fit, settings = "yes")

Data

A CSV example is included at inst/extdata/example_data.csv.

License

MIT.