Skip to contents

Bland-Altman agreement plot

Source: R/vis-clinical.R
plotBlandAltman.Rd

Draws the Bland-Altman difference-vs-mean scatter that blandAltman only computes: the bias (mean difference), the upper/lower limits of agreement (LoA), and shaded confidence bands for the bias and each LoA. The numerics come from blandAltman() so the plotted reference lines are single-sourced and cannot drift from the computed statistics.

Usage

plotBlandAltman(
  x,
  y,
  confidence = 0.95,
  proportional_bias = FALSE,
  units = NULL,
  colorblind = TRUE
)

Arguments

x, y

Numeric vectors of paired measurements (two methods, or two time points). Must be the same length with at least 2 pairs (validated by blandAltman()).

confidence

Confidence level for the limits of agreement and the shaded confidence bands (default 0.95).

proportional_bias

Logical. If TRUE, model the bias and LoA as linear in the mean (Bland-Altman regression): the bias line is the OLS fit of the difference on the mean with a shaded fitted-line confidence band, and the LoA are the fit \(\pm z \cdot s_{resid}\). Use when the difference scales with magnitude. Requires at least 3 pairs (the regression residual variance is undefined at n = 2). Default FALSE (constant bias/LoA).

units

Optional measurement unit appended to the axis labels.

colorblind

Logical; if TRUE (default) use the ecosystem colorblind-safe palette (physioPalette).

Value

A ggplot object.

References

Bland JM, Altman DG (1986). Statistical methods for assessing agreement between two methods of clinical measurement. Lancet, 327(8476), 307-310. doi:10.1016/S0140-6736(86)90837-8

See also

PhysioCore::blandAltman() for the underlying statistics.

Examples

# \donttest{
set.seed(1)
m1 <- rnorm(30, 50, 10)
m2 <- m1 + rnorm(30, 0, 3)
plotBlandAltman(m1, m2)

plotBlandAltman(m1, m2, proportional_bias = TRUE, units = "deg")

# }