• Question: How many times do you conduct a experiement for it to considered reliable?

    Asked by umera to Emma, Jimmy, Janet, Niall, Simon on 13 Mar 2013.
    • Photo: James Holloway

      James Holloway answered on 13 Mar 2013:

      In particle physics we have whats called the ‘5 sigma gold standard’. Which means you have to have enough data to say:
      “We are 99.9999% sure we have found a new discovery”

      But that is a very strict requirement and other sciences are not quite so anal.

      Typically your results have to be reproduceable by other scientists for them to be considered reliable.

    • Photo: Janet Daly

      Janet Daly answered on 13 Mar 2013:

      It depends a lot on what the experiment is measuring and how extreme the outcome is.

      For example, if you can vaccinate a group of people against influenza and then expose them to the virus and they don’t get sick but another group of people who were not vaccinated all get sick, then it’s fairly clear-cut and you don’t have to have lots of people in each group and only need to repeat the experiment once or twice. On the other hand, I’ve seen a study where they gave people who got naturally sick with influenza a drug that helped reduce how sick they got. One of the ways they measured how much the drug reduced the illness was by how many tissues people given the drug used to blow their noses compared to people who were not given the drug…this is not very precise as some people like to blow their nose more frequently than others anyway.

Comments