ISO/TR 24430:2003

Foreword
Introduction
1 Scope
2 Normative references
3 Terms and definitions
4 Requirements for an interlaboratory precision experiment
  4.1 General
  4.2 Personnel requirements
      4.2.1 The executive officer/project manager
      4.2.2 The statistician
      4.2.3 The operator
  4.3 Instructions for the interlaboratory trial
      4.3.1 Test method
      4.3.2 Measurement range
      4.3.3 Sampling and transport
      4.3.4 Number and labelling of samples
      4.3.5 Number of observations to be made
      4.3.6 Timing of the experiment
      4.3.7 Test report
      4.3.8 Comments
  4.4 Minimum conditions for interlaboratory experiments
      4.4.1 Participating laboratories
      4.4.2 The number and distribution of samples
5 Conducting the test
  5.1 Programme
  5.2 Pilot run
  5.3 Progress monitor
  5.4 Replacement samples
  5.5 Inspection of data
6 Computing the precision - The statistical analysis
  6.1 Data correction
      6.1.1 Redundant data
      6.1.2 Missing data
      6.1.3 Outliers
  6.2 Dependence of precision on level
  6.3 Calculation of precision
7 Reporting precision values
Annex A Description of the process

The purpose of ISO/TR 24430:2003 is:

• to give recommendations on how to introduce precision statements into an International Standard which specifies a test method for paints, varnishes, related materials and coatings derived from them;
• to outline the general principles to be applied when assessing precision in order to establish the procedures which will enable a quantitative estimate to be made;
• to outline the general principles under which interlaboratory trials should be conducted.

It may be applied to paints, varnishes and related products, to their raw materials in solid, liquid or powder form and to coatings derived from them.

It is concerned only with test methods which operate on a continuous scale to yield a single numerical figure as the test result. However, this single figure may be the outcome of a calculation from a set of measurements. The distribution of test results is required to be unimodal and is assumed to be normal. With non-Gaussian distributions other evaluation procedures will be necessary. It does not cover methods which yield discrete values such as go no-go tests or where a ranking scheme is in operation.