• Question: how accurate is your data

    Asked by lottie243 to David, Thiloka, Shonna, Shobhana, Ryan, Ross, Rebecca, Rachel, Patrick, Nina, MattyB, Matthew, Marianne, Lorena, Kate, Kaitlin, James, Ettie, Emmanuelle, Deepak, Anabel, Ambre, Alex, AlexAgrotis, Aina on 12 Jun 2019.
    • Photo: David Wilson

      David Wilson answered on 12 Jun 2019: last edited 12 Jun 2019 7:16 pm


      As accurate as we can make it. If we can be as accurate as possible with our measurements in experiments then our results should be accurate too.
      I read once that only Mathematics could be 100% accurate, this is because when you show something with a mathematical proof there is no question about it being correct. However, this isn’t the case in Biology. We might do an experiment 100 times and 99 times it does the same thing but 1 time it does something different. We could still be confident that our experiment is good but what we’re seeing in the 1 time it did something different is called variation and variation is at the centre of what biology is. For example, we are all humans with two arms, two legs and a head but we also know that we are all subtly unique and different, just look at your finger prints! This is why a biologist can never truly be 100% correct.
      Edit: since this generated a bit of discussion on twitter I went and dug out the book I read it in, I was quite surprised I could remember where I saw it! I read this as a teenager and I had a sort of epiphany that this made complete sense to me and explained why biology can never be 100% certain.
      It’s in a book called Fermat’s Last Theorum by Simon Singh, here’s the quote:
      “Proof is what lies at the heart of maths, and is what marks it out from other sciences. Other sciences have hypotheses that are tested against experimental evidence until they fail, and are overtaken by new hypotheses. In maths, absolute proof is the goal, and once something is proved, it is proved forever, with no room for change. “

    • Photo: Aina Roca Barcelo

      Aina Roca Barcelo answered on 12 Jun 2019:


      As David said, as accurate as we can. I usually use what we call “administrative data” which means that is data which is routinely collected for purposes other than research but that can be useful. For example, hospital data. Therefore, when you pick a dataset to use, you need to make sure you understand the extension of the data, how it was collected and its original purpose as this will condition how accurate or suitable it is for what you want to do.

    • Photo: Rebecca Moon

      Rebecca Moon answered on 12 Jun 2019:


      Accuracy of data is massively important to research. I work on studies involving hundreds of participants who are all having the same measurements taken, but these measurements might be made by different staff members including myself, research nurses or research assistants. We therefore have standardised protocols that we all follow to make sure we are doing the measurements in the same way, even for something that might seem obvious, such as measuring height. I take measurements of participants body shapes so head circumference, chest circumference, arm circumference and measurements of grip strength. When we do these, we repeat the measurements 3 times to make sure the results are similar and take an average. We also have to calibrate our scanning machines regularly using a “phantom”. This is a manmade “human” of known measurements that gets scanned every day to ensure that the scanner always comes up with the same answer.

    • Photo: Marianne King

      Marianne King answered on 13 Jun 2019:


      It’s difficult to know how accurate our data is as it’s hard for a biologist to prove something, because some organism somewhere will insist on behaving differently to all the other ones. For example, I work on a disease called Rett Syndrome and I want to make brain cells from patient stem cells. If only worked with one patient’s stem cells, there’d be a chance that anything I observed would only be the case for that patient. This is why I have four lots of patient stem cells, all from different patients. And for each experiment, I will do it three times. We also make sure that all our equipment is well calibrated and that the conditions we grow our cells in are the same. You want the only variable to be the cells themselves, because otherwise it’s much harder to explain any changes you might see.

    • Photo: Matthew Bareford

      Matthew Bareford answered on 14 Jun 2019:


      So long as my experiment is completed properly then my data is as accurate as possible for that specific reaction on that date at that time…..

      Because everything down to the slightest atom will be slightly different each time (electrons wont be in the same place, there might be more space between two atoms in a solution, etc.) then there will, always be slight variables.

      That is one of the joys of science!

    • Photo: Kaitlin Wade

      Kaitlin Wade answered on 14 Jun 2019:


      That’s a great question and one that we ask ourselves each day. I work with large collections of data that have been provided by participants of studies through things like them answering questionnaires, or coming into a clinic and getting measured for all sort of things like height and blood pressure, or even them giving blood (and other samples!) that provide us with information like genetics. All of this individually is a huge amount of information on someone’s lifestyle, health and behaviour but, together, it’s a wealth of data that helps us answer questions like “how does being obese change someone’s blood pressure?” etc. We have to make this data as accurate as possible so that it’s actually a good reflection of someone’s health at that time. Otherwise, we might get the wrong answers, say if someone has lied or remembered something incorrectly or stood on their tiptoes whilst getting their height measured. To make this as accurate as possible, we have to have a lot of faith in the participants themselves for things like questionnaires. For measures taken by clinicians or technicians (like blood pressure and height), we make sure that we take multiple readings at the same time and average them out, just in case one reading is particularly weird. But the interesting thing is, is that we can actually test how accurate something is, which helps us improve the measurement later in time, or whether someone has lied because this tells us a bit about who they are as a person. Everything is data and even inaccuracies can give us some information!

    • Photo: Nina Rzechorzek

      Nina Rzechorzek answered on 15 Jun 2019: last edited 15 Jun 2019 7:15 pm


      Brilliant question. First let’s be clear on what we mean by accuracy when it comes to data. A set of data can be accurate, or precise, or both, or neither! Given a set of data points from repeated measurements of the same quantity, the set is ‘precise’ if the values are close to each other, and the set is ‘accurate’ if their average is close to the ‘true value’ of the quantity being measured. For biologists, the main issue is that we rarely know what this ‘true value’ is (we’re usually trying to find it out!). When we extract data from a ‘sample’ of a biological population (e.g. human subjects, cells in a dish) we are making an assumption that our sample group represents the entire population – and this may not be the case. To be sure, we would have to sample the whole population (every human, and every cell – and that is clearly not feasible). So we have to be really careful about the sampling technique we are using, the size of our sample in relation to the question we are asking, and even more careful about how we interpret the results. As scientists, we are getting really good at extracting ‘precise’ data, but in some cases we could be way off target with respect to accuracy. It is possible to be very precisely wrong! That is why when we think we have discovered something important, we need to use many different types of experiment to demonstrate that the finding holds true by various different techniques – that way we can be as sure as we can that we are reporting the ‘truth’ (within a range that describes how confident we are in our results).
      As a scientist, the only thing I can be 100% certain of is that biological data will vary. As a clinician the only thing I can be 100% certain of is that not every patient will ‘read the textbook’….that’s why a treatment may not work as expected and you have to be ready with one, two, or multiple contingency plans.

      One of the main reasons I moved towards circadian research is because ‘time of day’ effects can have such a huge impact on biological phenomena – and pretty much everything we measure in the clinic, the lab, and out in the field….we need to figure out how this affects our data, how we can control for it, or how we can modify biological time in our favour!

Comments