SEARCH

  • Default
  • Title
  • Date
  • Random

Limits

The manufacturers limits are meant to give an idea of the range within which you should expect your labs mean to fall. They are probably set too wide to be used as routine control limits for an individual laboratory.

Certification of the CRM is based on a measurement campaign, so allocation of property values takes place on the basis of agreement among the independent measurement results and, not necessarily with direct traceability to Standard International units. Two of the ISO assumptions that allow this are; that there will be enough capable labs and that the results from each lab will be statistically compatible (ISO Guide 35 Clause 9.2.3 and Clause 10). To overcome the effects of this "inter-laboratory issue", and to achieve "a property value having satisfying uncertainty" (ISO Guide 35 Clause 10.2.2) CRM producers must raise the minimum number of laboratories involved (in practice to 15 laboratories per analyte, 8 samples to each) and scrutinize the data with the aid of outlier treatment techniques. This should lead to a very accurate measure for a given method; notwithstanding the underlying assumption that what the good inter-laboratory labs reported was accurate. However, an amount of bad data may have an effect, resulting in limits which may too broad for effective monitoring of a single laboratory or production process. This is a major concern to the CRM producer and it should be an issue (by simple extension) to the CRM consumer.

CRM consumers are advised to set their own limits, (normally within the CRM producers limits) based on their own data quality objectives and control measurements. For example, limits probably don't need to be as tight, for plant residues or for geochem grade levels; but they need to be tighter for high ore and concentrate grade levels (purely for financial considerations).

The CRM consumer may for various reasons choose to use the manufacturers limits to monitor the performance of a single laboratory.  However, if the lab is reporting accurately (no bias) then these limits are unlikely to adequately monitor precision 2SD failure's lower than 5% could be an issue (see Control Rules). If the lab has a bias, 2SD failures higher than 5% will become a problem.

The precision or bias issue thus identified might be due to either lab problems or CRM commutability and must be discussed with the laboratory manager. Commutability issues can be investigated using the matrix chemistry published in the CRM's certificate.  If commutability is not the cause of the greater or less than 5% 2SD failure rate, then the CRM's have done their job and the lab problem can be fixed.  That's what CRM's are for.

FAQS

  • 1