Abstract
Confidence intervals are usually constructed to assess the levels of precision in method validation studies. Frequently, these confidence intervals are constructed assuming that the random errors come from infinite populations. In practice, the random errors are selected from a finite population of known size. In these cases, the infinite population model provides confidence intervals that are too wide. A fixed effects model underestimates the measures of precision. In this article, confidence intervals for measures of precision are developed in an interlaboratory study with labs selected from a finite population of known size.