Exact Parametric Confidence Intervals For Bland-Altman Limits Of Agreement

Bland JM, Altman D. Statistical methods for assessing the concordance between two clinical measurement methods. Lanzette. 1986;327:307-10. Bland and Altman point out that two methods for measuring the same parameter (or property) should have a good correlation when a group of samples is selected, so the property to be determined varies greatly. A high correlation for two different methods designed to measure the same property could therefore in itself only be a sign that a widespread sample has been chosen. A high correlation does not necessarily mean that there is a good agreement between the two methods. Probability of coverage of 97.5% unilateral confidence interval for N = 10 Despite the positive results of previous research, detailed numerical assessments are presented to highlight the underlying disadvantages of the methods under the idea that the criteria of a bilateral confidence interval have an appropriate interpretation as the lower or upper confidence limit of a unilateral confidence interval. Basically, the simplicity and symmetry of an approximate confidence interval generally does not maintain the assumption of error rates on the same hand for the two breakpoints. For the planning of percentile studies, in order for the results to contribute to the confirmation of significant reference objectives, sample size methods are described to accurately estimate the interval between normal percentiles, according to the expected accuracy criteria, width and probability of safety. In order to improve the applicability of the exact interval approach and the corresponding sample size methods, computer codes are also presented to perform the necessary calculations.

with τ BAL = p N1/2 – t1 − α/2 (ν) b1/2 and τ BAU = z p N1/2 + t1 − α /2 (ν) b1/2. In the specific case of α = 0.05, the general expressions are reduced to the confidence intervals for both ends of the 95% correspondence limits taken into account in Bland and Altman [2]: for the derivation of the confidence intervals of θ. It is easy to see that T B = T* + p N1/2 and T MU = T* + p cN1/2. . . .