Taking multiple measurements

[turn to a video?]

As we have seen, there is no such thing as a single, precise and true piece of data. For any measurements we make, or value we look up, there is a range of possible values - this range is referred to as the “uncertainty”. With only this information, we know nothing about what is happening inside the limits of this range, and the true value could be anywhere within this range. We can design products to be robust and operate within this range of values, but this can lead to unnecessary cost or time by over engineering or over specifying requirements. If we knew a little more about what was happening within the range of uncertainty, we would have a more precise knowledge of the system we are measuring and potentially avoid wastage.

By taking multiple measurements we can reveal knowledge about the details of what is going on inside the uncertainty range. With more readings, we can build up a more accurate picture of the likelihood that the true value exists in more specific areas and predict, with a known degree of confidence, where we believe the true value might exist.

There are two cases where you would take multiple measurements, and they each have their own use and drawbacks. Perhaps it is the repeatability of a measurement you are looking to determine or maybe it is the reproducibility. You need to understand which one you are doing, as it will influence how you analyse and present the results.

Repeatability - if you repeat a measurement using the same instrument, it can tell you the scatter in the results due to the measurement equipment.

Reproducibility - If you repeat your experiment under exactly the same conditions, multiple times, it can identify the scatter in the results due to the process of conducting the experiment.

Consider an experiment using the combustion of fuel in a fire tornado:

Video of fire tornado, bit of text around it, same style as video of Ed.

Ball Rolling video