11.6 Summary

• To provide explanatory knowledge an experiment should be based on cause variables, not control variables.
• Random error is synonymous with noise and averaging over a sample reduces its influence. External noises arise outside the measurement system and can be handled by isolating the system or reducing the noise (for instance by a strict measurement procedure). Internal noises arise within the measurement system and are more difficult to reduce.
• Bias is synonymous with systematic error and is coupled to the measurement accuracy. It is normally handled through corrections or isolating the system. Randomization is a useful strategy for unknown systematic errors. It is important to consider systematic errors also when using relative data.
• Potential background factors in an experiment may be identified using the Ishikawa diagram (Chapter 10).
• It is useful to apply Ishikawa diagrams, process diagrams and input–output diagrams to the measurement system itself in order to develop strategies for avoiding sources of uncertainty.
• Accuracy is assessed by measuring a known standard. Precision is assessed by replicating measurements. Precision in raw data can be measured by the sample standard deviation. Precision in an averaged value is better measured by the standard error of the mean or a confidence interval.
• The data collection plan describes how the measurements should be carried out to obtain representative data of good quality. It should include a standard procedure for the measurements and a plan for regular stability checks. It could also include training of operators.