Journal of Statistics: Advances in Theory and ApplicationsAuthor's: Steven B. Kim
Pages: [119] - [137]
Received Date: April 21, 2019; Revised May 17, 2019
Submitted by:
DOI: http://dx.doi.org/10.18642/jsata_7100122049
Bland Altman analysis is a statistical method for assessing the degree
of agreement between two methods of measurement. In medical and health
sciences, it is a popular method because of its simple calculation and
visualization. Under normality assumption, the calculation is based on
two sufficient statistics 
and s, where is the sample mean of differences and s
is the sample standard deviation of the differences. The interval
is referred to as 95% limits of agreement (LOA)
in literature. In a seminar paper, Bland and Altman [2] interpreted
LOA as “If the differences are normally distributed, 95% of
differences will lie between these limits”. This interpretation
seems to be widely accepted, but there is a caveat because the
coverage probability of LOA is a random variable. In this article, we
demonstrate the sampling distribution of its coverage probability by
simulation, and we discuss an alternative choice for the critical
value. In addition, using simulation, we perform sample size
calculation which satisfies a specified condition for the sampling
distribution of coverage probability.
Bland Altman analysis, limits of agreement, prediction intervals, tolerance intervals, sample size calculation.
