Articles Blog

Science Survival Kit: Data-driven Research

Science Survival Kit: Data-driven Research


We have already seen how the classical scientific
method looks like. It is also called hypothesis-driven research. We start in the beginning with an observation
and a question, we search the literature, we propose a hypothesis that we will test, and then we
obtain the results that will tell weather the hypothesis was true or false. This clear and streamlined model is often
not what happens in real life. In real life, there are usually many problems
along the way. Sometimes we might not be able to propose a
hypothesis, because we will not have enough background knowledge we need to do more observations and more bibliography. Other times, the experiments might not work,
we might have to abandon our experimental design and seek for new solutions to test the hypothesis. Due to the lack of time, resources or skills,
we might not obtain enough data to reject or confirm the hypothesis. This means we will have to repeat the experiment to obtain more data Occasionally, we might be testing our hypothesis,
and we might see an interesting phenomenon that we have never noticed before. This unexpected observation might be more
valuable than the initial project that we were studying and maybe we want to continue studying it further. so you can see, the scientific method that
looks very linear is in fact, in real life not linear at all. What might happen one day is that we are studying
a phenomenon that nobody knows about we might be actually the first ones that will be studying it. What we have to do then is to resort to exploratory
data analysis. This means that we will try to aquire as much data as possible about that specific phenomenon, and then try to analyze the data to draw conclusions. This is called data-driven research. Data-driven research is especially important today, when we have more and more high-throughput devices which allow us to get a lot of quantitative data about the different phenomenon we are trying to study. Analysing such “big data” is very challenging but careful examination of such data sets can provide us a lot of new information we might see new patterns and different correlations between variables. This should remind you of something! Once we have the correlation between two
variables, we can test for causality. We can try to examin if one of them
is a cause and the other one a consequence. We can form a hypothesis, that states: if we change the independent variable (this will be ourpossible cause), then this will happen to the dependent variable
(the possible consequence). So in this way you can see that the data-driven research can drive the hypothesis-driven research. Both of them should work together to advance
science and extend the frontiers of our knowledge.

Leave a Reply

Your email address will not be published. Required fields are marked *