ISO/TC 45/SC 2/WG 4 - Application of statistical methods

ISO 19983:2017 provides guidelines and specifies requirements for estimating the precision of rubber test methods by means of interlaboratory test programmes based on the procedures given in ISO 5725 (all parts).

ISO/TR 12134:2010 provides guidance to scientists, engineers and technicians, working in the field of rubber materials and products, to supplement ISO/IEC Guide 98‑3 and to provide additional guidance in situations where functional relationships between input quantities (such as temperature, strain rate and time) and derived output quantities (such as tensile strength and compression set) are unknown and where no other guidance is available. The Technical Report provides a summary of the classical approach that is taken in the preparation of uncertainty budgets and provides a list of selected test methods and, for each, an indication of the factors that will make a contribution to the uncertainty budget.It also discusses how “non-functional” factors can be taken into account in the classical approach.

ISO 19003:2006 provides guidance on the application of statistics to rubber testing. It is intended not to conflict with or replace existing International Standards covering basic statistical techniques, but rather to complement them and provide examples of those techniques applied to particular rubber testing situations.

ISO/TR 9272:2005 presents guidelines for determining, by means of interlaboratory test programmes (ITPs), precision for test method standards used in the rubber manufacturing and the carbon black industries. It uses the basic one-way analysis of variance calculation algorithms of ISO 5725 and as many of the terms and definitions of ISO 5725 as possible that do not conflict with the past history and procedures for precision determination in these two industries. Although bias is not determined in this Technical Report, it is an essential concept in understanding precision determination. The ISO 5725 concepts of accuracy and trueness are not determined in this Technical Report. Two precision determination methods are given that are described as 'robust' statistical procedures that attempt to eliminate or substantially decrease the influence of outliers. The first is a 'level 1 precision' procedure intended for all test methods in the rubber manufacturing industry and the second is a specific variation of the general precision procedure, designated 'level 2 precision', that applies to carbon black testing. Both of these use the same uniform level experimental design and the Mandel h and k statistics to review the precision database for potential outliers. However, they use slight modifications in the procedure for rejecting incompatible data values as outliers. The 'level 2 precision' procedure is specific as to the number of replicates per database cell or material-laboratory combination.

ISO 19004:2004 describes how test sensitivity can be determined for test methods used to measure typical physical and chemical properties of rubber and rubber products. It is also applicable to tests used to measure the properties of carbon black fillers. Test sensitivity is defined as the ratio of the discrimination power of a test method for the fundamental property to be determined to the measurement error expressed as a standard deviation. It is frequently described as the 'signal-to-noise ratio'. The standard does not address the topic of sensitivity in the context of threshold (i.e. minimum) detection limits in the determination of very low or trace constituent levels. ISO 19004:2004 is based, with permission of ASTM, on ASTM D 6600-00, Standard Practice for Evaluating Test Sensitivity for Rubber Test Methods, copyright ASTM.

Presents guidelines for evaluating the accuracy and bias of ISO/TC 54 test method results. These guidelines are additional to those in ISO 5725 and ISO/TR 9272, and give the format for expressing accuracy and bias. Is limited to test method standards that have test results expressed in terms of quantitative continuous variable or have test results which are quantifiable in absolute terms.

Describes a method for obtaining confidence intervals for the unknown precision values. On the assumption of normal distribution the confidence intervals are stated with a preselected error probability of 0,10.