Address:
|
100 Bureau Dr.
, Stop 8443 Gaithersburg, MD, 20899-8443, United States
|
Abstract:
|
The model supplementary comparison analyzed in this paper is from Consultative Committee on Photometry and Radiometry (CCPR) Supplementary comparison S2, a portion of which is presented briefly in Appendix A of the Draft B report, and details an optional method not used to calculate the official results of the comparison. The analysis details measurements of aperture areas involving 9 laboratories using 8 distinct artifacts, each having a different nominal value. The apertures are made of two materials (copper and aluminum bronze), in two sizes (25 mm and 6 mm), using two finishing manufacturing methods (diamond turned and conventional), deployed to create apertures of two edge types (knife edge and cylindrical). Artifacts are considered to be exchanged according to a star design, being sent from pilot laboratory to a participating laboratory and back. Each laboratory reports a mean value of several measurements over a short time period, accompanied by estimates of type A and type B uncertainties. The pilot laboratory is responsible for the ongoing stability checking of artifacts. The evaluation assumes all artifacts to be different with distinct true measured values. Outliers are tested by graphical method using h and k Mandel's statistics. Individual laboratory estimates are compared with a comparison reference value (RV), not required for supplementary comparisons, but presented for convenience. The reference value is computed here as a maximum likelihood estimate of the consensus mean across all laboratories. The evaluation also includes the investigation of effects attributable to different measurement methods, performed on three artifacts that were measured by several laboratories using both contact and non-contact method. This supplementary comparison represents a first pass in the area. The pilot laboratory proposed the use of an optional second method to compare the reproducibility values of the two different methods. More complex experimental design would enable the study of other influential factors such as differing manufactures or measurement positioning schemes.
|