Statistical Engineering Division, ITL
The Statistical Engineering Division supports the Standard Reference Materials Program and the other NIST laboratories by collaborating directly with chemists and other scientists engaged in the certification of Standard Reference Materials (SRMs). SRMs are artifacts or chemical compositions that are manufactured according to strict specifications and certified by NIST for one or more quantities of interest. SRMs are a primary vehicle for disseminating measurement technology to industry.
Development of a new SRM typically takes about five years and encompasses: 1) validation of a measurement method; 2) design of a prototype; 3) stability testing; 4) study of measurement error; 5) certification uncertainty analysis. Statisticians advise on the design and analysis of experiments at all phases; develop estimation methods for data from different analytical methods; help reconcile interlaboratory differences; and combine all information to produce a certified value and statement of uncertainty.
In 1997, thirty-five new SRMs were assigned to division staff; fifty SRM certifications were completed; and approximately fifty SRMs are still in some phase of development. The largest number of SRMs, far and away, came from the Chemical Science Technology Laboratory (CSTL), but SRMs from the other NIST laboratories covered a variety of applications, for example: MEL (e.g. sinusoidal roughness); MSEL (e.g. Knoop hardness); EEEL (e.g. resistivity of silicon wafers) PL (e.g. optical density filters); BFRL (e.g. thermal resistance of fibrous glass insulation).
Tracking such a large number of SRMs has always been challenging because the responsibility for the status of the certification process often shifts back and forth between scientist and statistician as follow-on data are taken and analyzed or more analytes are added to the certification process. This year an internal web page was put on-line that allows NIST staff to determine the status of SRMS in SED.
The large workload has led the division to consider more efficient methods for handling the statistical design and analyses of SRMs. As a start, a common protocol for certifying gas mixture standards was developed in collaboration with chemists from CSTL. From this protocol, chemists developed a spread-sheet for handling certifications and now require only occasional assistance from statisticians for fifty or more issues per year.
Date created: 7/20/2001