An Evaluation of Methods for Assessing the Competence of Laboratories Based on Performance in the GeoPT Proficiency Testing Scheme. Issue 2 (1st February 2019)

• Record Type: Journal Article
• Authors: Potts, Philip J.; Webb, Peter C.
• Is Part Of: Geostandards and geoanalytical research. Volume 43:Issue 2(2019)
• Page Start: 217
• Page End: 229
• Publication Date: 2019-02-01
• Subjects: reference material certification -- laboratory competence -- laboratory selection -- GeoPT -- proficiency testing -- USGS SyMP‐1 -- syenite -- CGL 006; Analytical geochemistry -- Periodicals; Géochimie analytique -- Périodiques; 551.9
• DOI: 10.1111/ggr.12252
• Languages: English
• ISSNs: 1639-4488

Abstract

Abstract : The original certification protocol, published by the International Association of Geoanalysts in 2003, specified that the competence of laboratories selected as competent to contribute certification measurements should be evaluated from their performance in the Geo PT proficiency testing programme. Round 39 of the IAG Geo PT proficiency testing programme provided an opportunity to examine four methods of evaluating laboratory competence based largely on the use of proficiency testing z ‐scores as performance indicators. This opportunity arose because two test materials were co‐analysed by participating laboratories in this round: a syenite, SyMP‐1, supplied by the USGS, and an established CRM, the nepheline syenite, CGL 006. The performance of laboratories was assessed in four ways; in each case, consensus values and their uncertainties as derived from selective data sets of competent laboratories were compared with results derived from the routine Geo PT data assessment, involving all submitted measurements. An overall comparison of results showed no significant statistical differences in either consensus values or uncertainties between these data sets. This conclusion was unexpected and calls into question the widely held assumption that 'better' consensus data would be obtained from a subset of laboratories judged to be competent on the basis of proficiency testing performance indicators. Key Points: Four different ways were investigated for selecting the best performing laboratories from Geo PT data sets. Consensus values and uncertainties from these selected data subsets were statistically indistinguishable from equivalent results from the full Geo PT data set. The assumption that better quality data can be obtained from a subset of superior performing laboratories in the Geo PT programme is wrong.