11 .Precision and Bias 11.1 Interla

11 .Precision and Bias 11.1 Interlaboratory tests. Tests were conducted in 1980 with eight laboratories evaluating four fabrics under washing and drying conditions 1-III-A and 1-IV A of AATCC Method 124. The analysis of variance technique was judged not to be applicable to this data set because its distribution was not normal, and because of the limited and discontinuous scale of replica grades. The data were analyzed by calculating expected laboratory test results from the distribution of individual specimen grades. This analysis has been deposited for reference in the RA61 committee files. 11.2 Observer repeatability. From the data it was determined that single observers rated three specimens on the following frequency: 3 specimens to same replica ..........0.55 2 specimens to same replica and one different ..............0.40 3 specimens different.....................0.05Only rarely did the separation in specimen grades exceed the next replica step.This is indicative of the high degree of repeatability in observer rating of smoothnessappearance. 11.3 Laboratory test result distribution 加ithin-laboratory repeatability). From the observed grade distribution, a distribution of laboratory test results was calculated for each replica level with halfgrades included. Precision over the whole SA replica range was improved. 11.4 Precision. From the frequency distribution of laboratory test results, a calculation was made of the critical difference,D, between two laboratory test results.With laboratories at the same level: CriticalDifferenceD>0.17D>0.25Confidence LevelP >_ 0.95P >0.99When two or more laboratories wish to compare test results, it is recommended that laboratory level be established between them prior to commencing test comparisons. Fabrics of known history and performance may be used for this purpose. Differences between laboratory test resups (on the same fabric, under the same washing and drying conditions) equal to