Grouped and stratified Bland-Altman analysis

Why group the analysis?

Agreement can differ across biological or operational subgroups.
ggBA supports grouped summaries and faceted plots to inspect those patterns.

library(dplyr)
#> 
#> Attaching package: 'dplyr'
#> The following objects are masked from 'package:stats':
#> 
#>     filter, lag
#> The following objects are masked from 'package:base':
#> 
#>     intersect, setdiff, setequal, union
library(knitr)
library(tidyr)
library(ggBA)

Build paired data once

tbl <- temperature |>
  pivot_wider(names_from = method, values_from = temperature)

Grouped statistics by treatment

treatment_stats <- ba_stat(
  data = tbl,
  var1 = infrared,
  var2 = rectal,
  group = treatment
)

treatment_stats |>
  filter(parameter %in% c("bias", "lloa", "uloa")) |>
  tidyr::pivot_wider(names_from = parameter, values_from = value) |>
  arrange(treatment) |>
  kable(digits = 3)

treatment	n	bias	lloa	uloa
healthy	75	0.202	-2.511	2.915
vehicle	75	0.038	-3.226	3.303
low dose	75	0.035	-3.157	3.228
mid dose	75	0.634	-2.446	3.715
high dose	75	0.173	-3.153	3.498
SoC	75	0.319	-2.504	3.142

This table highlights subgroup-level shifts in bias and spread.

Faceted Bland-Altman plots

ba_plot(
  data = tbl,
  var1 = infrared,
  var2 = rectal,
  group = treatment,
  colour = visit,
  title = "Agreement by treatment group",
  caption = "Each panel has treatment-specific Bland-Altman summary lines."
)

Optional: log-scale grouped analysis

If multiplicative differences are more relevant, add transform = "log":

ba_stat(
  data = tbl,
  var1 = infrared,
  var2 = rectal,
  group = treatment,
  transform = "log"
) |>
  filter(parameter == "bias") |>
  mutate(geom_ratio = exp(value)) |>
  select(treatment, geom_ratio) |>
  arrange(treatment) |>
  kable(digits = 3)

treatment	geom_ratio
healthy	1.006
vehicle	1.001
low dose	1.001
mid dose	1.017
high dose	1.005
SoC	1.009