Ask Mallard

Bland–Altman (method comparison)

§7.2
Rev 4.6 · reviewed Feb 2026
Bland–Altman — method comparison
Assesses whether two measurement methods agree well enough to be interchangeable. Correlation is the wrong tool here; agreement is. Paste paired measurements in the same order.
Source formula: Bland JM, Altman DG. Lancet 1986;327:307–10.
Method A
Method B
bias -0.3+1.96 SD 5.8−1.96 SD -6.4mean of A & B →
Pairs (n)10
Bias (mean difference)-0.30
Limits of agreement-6.43 to 5.83
* Limits = bias ± 1.96 SD of differences. Bland JM, Altman DG. Lancet 1986;327:307–10.
Method A reads on average 0.30 units lower than B, and 95% of differences fall between -6.43 and 5.83. Whether that's acceptable is a clinical judgment — compare the limits to the largest difference that wouldn't change management, not to a p-value.
Understand what this is measuring