Experimental Design – Are You Asking The Right Questions?
Many a scientist will be able to tell you that hypothesis testing is what science is all about. You uncover a mystery or a problem, based on your knowledge and past research, you formulate a hypothesis to explain what you are seeing and then you test this hypothesis to get an answer.
It is in this testing phase that experimental design plays a key role. Design it well, and you will have confidence in your conclusions. Yet, if your experiments are not designed well, you could end up proving nothing and simply wasting your time.
So when it comes to experimental design – these are the questions you should be asking to cover your entire basis.
Did I do enough research?
There is a reason why most studies involve a detailed and extensive literature review before the experimental phase starts. This review allows you to, first of all, get enough background on the problem to be able to formulate the right research questions, and secondly, it allows you to assess whether your proposed experiments have been done before.
One of the key facets of science is that it should be novel, thus adding to the body of information already available. If you do not do the proper research beforehand, you may end up simply repeating someone else’s work.
What are the key influencing variables?
Each problem you encounter in life, or work will have key variables. These are the things that have the potential to directly influence the outcome of an experiment. As part of experimental design, you will need to figure out what these variables are, and how they have the potential to influence your experiment’s outcome.
Most science recognizes three main types of variables – dependent, independent and controlled variables. The independent variable is the one that you will ultimately manipulate in the lab or field to observe the associated response in the dependent variable. The controlled variable is the one that is kept constant so that it theoretically does not influence the dependent variable.
For example – if your research question is “How does temperature influence the moistness of a cake?” you can evaluate the main factors that would influence the moistness of the cake once it is baked. These would be the cake mix, the temperature you bake it at and the time it spends in the oven. You can standardize the experiment by using a pre-packed cake mix, and the same oven and the same amount of oven time for every set of the experiment – these are your controlled variables.
Naturally, the moistness of the cake would be your dependent variable – the one that you are observing, while the temperature you are baking the cake at would be your independent variable. Now all you have to do is decide what the settings of your independent variable should be – keeping in mind that the cake would still need to be completely baked by the end of a set period of time.
If the recipe recommends that the cake be baked for 45 minutes at 150ºC, you could try baking it at a range of temperatures going from 140ºC – 180ºC, and observe the results.
Missing a key influencing variable may have a huge effect on the outcome of your experiments, and could discount your findings very quickly. It is therefore very important to spend some time thinking and researching each scenario.
Is the experiment truly randomized?
It is assumed that there are always variables that cannot be accounted for in an experiment, either because we are not aware of them, or because they are difficult to measure or observe. For example, if we would like to determine if a new anti-depressant is effective, we would assign patients enrolled for the trial to one of two groups – the target group receiving the new antidepressant and a control group that will receive only a placebo.
Now, even though all the patients enrolled in the trial will have been diagnosed with depression, there are factors such as personality, individual variation and personal circumstance that cannot be controlled for. So, to randomize these factors between the control and target groups, patients are assigned randomly to each group.
Although randomization is important, its effects are only really seen with large groups as the effect of chance are diluted, and in many small studies, this is noted as being a problem.
Now you have an idea about what are the right questions you should ask before you proceed with your experimental design. Once you have your research ready, we recommend you to consult an expert editor to have a final and thorough look at your manuscript. This will help you get a well polished document ready for publishing without those embarrassing mistakes.
For more information contact firstname.lastname@example.org