Bland-Altman Analysis for Method Agreement

Performs a Bland-Altman analysis comparing two measurement methods or two time points. Computes the bias (mean difference), limits of agreement, and confidence interval for the bias.

Usage

blandAltman(x, y, confidence = 0.95)

Arguments

x

Numeric vector of measurements from method/time 1.

y

Numeric vector of measurements from method/time 2.

confidence

Numeric; confidence level for limits of agreement (default 0.95).

Value

A list with components:

bias

Mean difference (x - y)

lower_loa

Lower limit of agreement

upper_loa

Upper limit of agreement

sd_diff

Standard deviation of differences

ci_bias

Two-element vector with lower and upper CI for the bias

Details

The Bland-Altman method assesses agreement between two measurements by plotting their difference against their mean. The limits of agreement are: $$LoA = \bar{d} \pm z \times SD_d$$ where \(\bar{d}\) is the mean difference (bias) and \(SD_d\) is the SD of differences.

References

Bland JM, Altman DG (1986). Statistical methods for assessing agreement between two methods of clinical measurement. Lancet, 327(8476), 307-310.

See also

icc() for intraclass correlation reliability analysis, benchmarkAgreement() for benchmark validation agreement metrics, and plotBlandAltman() in the PhysioMoCap package to visualize the agreement.

Examples

set.seed(42)
method1 <- rnorm(30, mean = 50, sd = 10)
method2 <- method1 + rnorm(30, mean = 0, sd = 3)
result <- blandAltman(method1, method2)
result$bias
#> [1] 0.3657255
result$lower_loa
#> [1] -5.808493
result$upper_loa
#> [1] 6.539944