ISO Guide 35:2006 Reference materials - General and statistical principles for certification(修订中)
This Guide gives
statistical principles to assist in the understanding and development of valid
methods to assign values to properties of a reference material, including the
evaluation of their associated uncertainty, and establish their metrological
traceability. Reference materials (RMs) that undergo all steps described in
this Guide are usually accompanied by a certificate and called a certified
reference material (CRM). This Guide will be useful in establishing the full
potential of CRMs as aids to ensure the comparability, accuracy and
compatibility of measurement results on a national or international scale. In order to be comparable across borders and
over time, measurements need be traceable to appropriate and stated references.
CRMs play a key role in implementing the concept of traceability of measurement
results in chemistry, biology and physics among other sciences dealing with
materials and/or samples. Laboratories use these CRMs as readily accessible
measurement standards to establish traceability of their measurement results to
international standards. The property values carried by a CRM can be made
traceable to SI units or other internationally agreed units during production.
This Guide explains how methods can be developed that will lead to well
established property values, which are made traceable to appropriate stated
references. It covers a very wide range of materials (matrices), ranging from
gas mixtures to biological materials, and a very wide range of properties,
ranging from chemical composition to physical and immunoassay properties. The approaches described in this Guide are
not intended to be comprehensive in every respect of the production of an RM
and the establishment of its property values, including the associated
uncertainties. The approaches given in this Guide can be regarded as mainstream
approaches for the production and value assignment of large groups of RMs, but
appropriate amendments can be needed in a particular case. The statistical
methods described exemplify the outlined approaches, and assume, e.g., normally
distributed data. In particular when
data are definitely not normally distributed, other statistical methods may be
preferred to obtain valid property values and associated uncertainties. This
Guide describes in general terms the design of projects to produce a CRM.