• Question: How do you know that your data is accurate? How much data do you need to be sure in your conclusions?

    Asked by Sophie_A to Stewart, Miriam, Laura, David on 16 Nov 2018. This question was also asked by siena.
    • Photo: David Ho

      David Ho answered on 16 Nov 2018:


      In science we can never be 100% sure that our data is fully accurate. There’s always a possibility it could be a coincidence. But the more data we get, the smaller that probability gets. Scientists use mathematical tools to work out a number that tells you how likely it is that the results are due to chance. In particle physics, we normally require what is called a “5 sigma” confidence to announce a new discovery. That means the chance of our results being due to a coincidence is 1 in 3.5 million!

    • Photo: Miriam Hogg

      Miriam Hogg answered on 16 Nov 2018:


      We normally have ways of finding out how accurate the data is by knowing how accurate the inputted data is and finding the errors. In science papers you will often seen something like “the Universe is 13.8 billions years old +/- 0.5” for example. The +/- part is looking at the accuracy of our measurements and data and working out how far out it could be from the truth. So we think the Universe is 13.8 billion years but it could be as high as 14.3 billion years or as low as 13.3 billion years based on the accuracy of our data. (note: i’m not sure if the errors are that big for the age of the universe, it was just to help explain the idea 🙂 )
      As David already explained we also have ‘confidence’ calculations that are quite complicated and work out how likely something is to be a coincidence.

      My work is often simulating stars and planets so I have data from real ones to use and to compare to to see if my results make sense.

Comments