Analysis of Components of Variation

Abstract

An attempt was made to characterize quantitatively the several components of variation existing in the determination of total lung capacity (TLC) by a radiological method. An analysis of variance indicated that over 94% of the total variation could be attributed to film differences, the remaining 6% being a consequence of interobserver and intraobserver variation and the inherent random variability. It was shown that a relative error of about 5.8% could be expected with one observer making one determination for each set of films. An error of this magnitude compares favorably with other methods for determining TLC. It was, therefore, concluded that TLC determined by this method is repeatable and consequently can be used in large-scale epidemiological studies.