Many certifying agencies and some reference material data compilers regularly assign quality identifiers to their values, e.g. "certified values", "recommended values", "informational values", etc. Where an agency or a compiler has used these characterizations, we will attempt to faithfully reproduce them in our tables. In cases where these are not provided, Potts and Kane, 1992 (92POT 1), have proposed a framework, detailed in the following quotation, that we have adopted and are implementing:

Proposal for Revising Terminology

Reference materials that are characterized by "certifying bodies" using a rigorous analytical test program with a well-constrained assessment of results would be unaffected by the present proposals. The current International Organization for Standardization (ISO) terminology for "certified value" (and uncertified value) are defined in an unambiguous manner and already command respect from users in the confidence that may be placed in such data relative to specified confidence limits.

However, for existing silicate reference samples which are not characterized by formal certifying bodies, it is proposed that terminology be adopted which categorizes values into three confidence bands. Implicit in this proposal is an assumption that the issuing organizations must estimate reliably the uncertainty in the best estimate value, accounting for both random and systematic errors. It is also proposed that the uncertainty limits applicable to each confidence band should take into account the concentration of the analyte to reflect the variation in precision attainable by modern analytical techniques. To this end, we have adopted the scheme of Xie et al. (89XIE 1) and divided values into six concentration bands, as listed in Table 1. The proposed confidence limit boundaries are also listed in this Table. For values that are derived by interlaboratory testing, the following terminology is proposed to describe data that fall into each of the confidence bands:

Category 1 data: "certified value equivalent" Category 2 data: "recommended value" Category 3 data: "provisional value"

Further criteria in adopting these terms are as follows:

Category 1 data: certified value equivalent
"Certified value equivalent" is proposed for values that have been characterized with the greatest confidence, meeting the relevant standard deviation limits summarized in Table 1. It is assumed that estimates of standard deviation have been derived from a "representative" data set, comprising determinations by at least 6 independent laboratories, which includes data obtained from at least two independent, well-established techniques. For certain elements that can normally be analyzed by using a single technique (e.g. ferrous iron, loss-on-ignition), data are considered to have satisfied criteria in Table 1 for use of the term "certified value equivalent," provided that results are contributed from at least 10 independent laboratories. The expectation is that values placed in this category will not require revision during the lifetime of the reference material. The proposed limits are not entirely arbitrary and have been deliberately set high for two reasons. The first is in recognition of the high current expectations in accuracy and precision of contemporary analytical techniques, and the second is to compensate for the observation that estimates of confidence limits derived from co-operative trials may sometimes be over-optimistic (87AMC 1). For many current reference materials, it is expected that a significant proportion of the major elements may satisfy the above criteria, but only a smaller proportion of the minor and trace element data.

Category 2: recommended values
This category is reserved for values that have been characterized to a lower degree of confidence as listed in Table 1. Again, it is assumed that estimates of standard deviation have been derived from a "representative" data set, comprising determinations by at least 6 independent laboratories which include data obtained from at least two independent techniques (there being no significant systematic bias between techniques) or, alternatively, at least 10 independent laboratories using the same technique. Although values placed in this category may be subject to slight revision during the lifetime of the reference sample, it is expected that such data are characterized with sufficient confidence to justify their use in the calibration of instruments for geoanalytical measurements or in an assessment of accuracy, provided that such use is based on at least 5 reliable reference materials.

Category 3: provisional values
Values that satisfy neither of the above two categories but are still considered to be analytically informative are placed in this group. It is appropriate to include values in this category that are derived from small data sets or that have been derived by applying subjective judgment to a significantly incoherent set of contributed results.

The relationship between terminology described above and that used for data derived from certified reference material programs is summarized in Table 2.

Table 1: Concentration bands and proposed confidence limit categories based on the relative standard deviation (RSD % one sigma)

Concentration band of analyte	Certified Value Equivalent (E) Category 1 confidence limits (one sigma, RSD)	Recommended Value (R) Category 2 confidence limits (one sigma, RSD)	Provisional Value (P) Category 3 confidence limits (one sigma, RSD)
10 - 100 %	< 1 %	< 5 %	>= 5 %
1 - 10 %	< 2.5 %	< 10 %	>= 10 %
0.1 - 1 %	< 5 %	< 15 %	>= 15 %
10 - 1000 ppm	< 5 %	< 20 %	>= 20 %
0.1 - 10 ppm	< 7.5 %	< 30 %	>= 30 %
< 0.1 ppm	< 10 %	< 40 %	>= 40 %
Analytical criteria (see also text)	(i) At least 6 independent values obtained from at least two independent well-established techniques	(ii)As(i) or at least 10 independent values obtained from at least one well-established technique	(iii) Data not satisfying either of the other two categories but still considered to be analytically informative

Table 2: Comparison in terminology used in certification programs (ISO definitions from 78ISO 1) and that proposed for interlaboratory co-operative testing

Origin of data:	Data derived from a "certification program"	Data derived from co-operative interlaboratory testing
Criteria for categorization:	Values categorized according to the relevant ISO definitions	Values categorized according to best estimates of RSDs in relation to concentration band as listed in Table 1
Category 1	Certified Value (ISO definition)	Certified Value Equivalent
Category 2	non-certified value (ISO definition)	Recommended Value
Category 3	magnitude (or approximate composition)	Provisional Value

Note that the confidence status of "non-certified value" and "recommended value" are not equivalent.

L O S   A L A M O S   N A T I O N A L   L A B O R A T O R Y
Operated by the University of California for the US Department of Energy