plot.bland_altman

plot.bland_altman(result, clock, *, age_col='age')

Difference against mean, with bias and 95% limits of agreement.

A correlation coefficient cannot show that a clock’s error depends on age. This can, and age-dependent error is the failure mode that makes a single MedAE meaningless: a clock accurate at 40 and eight years out at 80 reports the same summary statistic as one uniformly four years out.