Measurements by the first method.
Measurements by the second method, paired with x.
See BlandAltmanOptions.
See BlandAltmanResult.
TypeError if x and y differ in length; RangeError for fewer than 2 complete pairs, infinite values, a
two that is not positive and finite, a mode other than 1 or 2, a level outside [0, 1] or an unknown
ciMethod.
const x = [10.1, 12.3, 9.8, 15.2, 11.1, 13.4, 14.8, 9.9, 12.0, 16.3];
const y = [10.6, 12.0, 10.4, 15.9, 11.0, 14.1, 15.3, 10.3, 12.9, 16.8];
const r = blandAltman(x, y);
r.meanDiff; // -0.44000000000000006
[r.lowerLimit, r.upperLimit]; // [-1.1628122854517629, 0.28281228545176262]
r.upperLimitConfInt; // [-0.17412106328886995, 0.73974563419239514]
blandAltman(x, y, { ciMethod: "exact" }).upperLimitConfInt; // [-0.011029811389107025, 0.961726537758047]
Bland-Altman analysis of two methods of measurement: the mean difference, the limits of agreement and their confidence intervals. Equivalent to
BlandAltmanLeh::bland.altman.stats()in R.With d the differences of the complete pairs (pairs with
NaNin either sample are dropped), d̄ their mean and s their standard deviation, the limits of agreement ared̄ ± two s. The mean difference has the t intervald̄ ± t s/sqrt(n)with t quantiles on n - 1 degrees of freedom. The interval of each limit is, by default, BlandAltmanLeh'slimit ± t sqrt(3 s²/n)(Bland and Altman 1986);ciMethod: "exact"gives the exact interval (Carkeet 2015):d̄ + s q/sqrt(n)for the upper limit with q the quantiles (1 ∓ confLevel)/2 of the noncentral t distribution with n - 1 degrees of freedom and noncentralitytwo sqrt(n), and its mirror image for the lower limit.Unlike R, fewer than 2 complete pairs, infinite values and a level outside [0, 1] are errors (R gives NA or NaN limits);
confLevel0 gives intervals of width 0 and 1 infinite ones, as R. Identical methods (s = 0) give every limit and interval equal to d̄.