Experiment!: Planning, Implementing and Interpreting

11.4 Measurement System Analysis

In the preceding section we focused on identifying the sources of uncertainty. We also discussed the ways in which different types of uncertainties can be handled. The time has now come to quantify how great the uncertainty is. Before we have determined the accuracy and precision of our measurement system we do not know if it is sufficient for our purpose.

As the precision reflects a random variation between measurements, it is obtained by replicating the same measurement many times. These replicates should be taken at different times to include as many sources of noise as possible. External noises, especially, tend to vary over time. They arise through small variations in the environment, difficulties in repeating experimental settings and so on. This means that consecutive measurements tend to be more similar to each other than measurements made at different occasions, so they underestimate the effects of external noises. One of the better ways to assess the precision is to run regular stability checks during the measurements – a procedure that will be described further at the end of the next section.

Let us say that we are interested in the precision of the measurements with the bathroom scales above. This is obtained by weighing the same person at a number of different occasions. The person's weight will of course fluctuate from day to day depending on food intake and so on, so it may be tempting to use a constant weight instead of a person. ...