Reference interval (RI) is a statistical inference which holds profound interest for diagnostic laboratories and assay manufacturers. Estimating population RI by relying on a single set of random samples holds multiple challenges, e.g. large required sample size and results dependent on the choice of method, and even dilemmas, such as how to interpret the confidence interval of an interval or the impact of outliers on interval estimation.
The Bayesian posterior distribution appears especially well suited for an RI inquiry, offering all the information usually sought while avoiding many challenges and difficulties. Even the often arduous Bayesian computation is relatively easy in an RI study setting, as the relationship between the data and the prior distribution is often simple and straightforward.
In this study, we compare the Bayesian and conventional approaches in determining RI using simulated data. Specific focus is given to 1.) the effect of sample size on the precision of RI estimates, and 2.) the impact of outliers on the accuracy of estimation.
|
ASA Meetings Department
732 North Washington Street, Alexandria, VA 22314
(703) 684-1221 • meetings@amstat.org
Copyright © American Statistical Association.