This editorial gives an overview of experimental design. Experimental design is the outline of how the researchers plan to collect data. During an experiment, there's one main factor-- the treatment. It's what researchers administer to the experimental units. It's their explanatory variable. It's the thing that we're trying to determine whether or not it has an effect.
In this short example here of a group of Alzheimer's patients being divided into two groups where one group is given a new drug, and the other group is given a sugar pill, the new drug is the treatment.
Now, when we're designing the treatment in the rest of our experiment, there's three main factors that are part of a great experiment. It's careful about control. It uses randomization. And there's a lot of replication.
Now, control is when you eliminate the influence of other variables. By eliminating the influence of other variables, the path from cause to effect is really clear. And we can determine whether or not the treatment was effective or had the intended outcome.
With randomization, that's when we're using chance to assign experimental units to the treatment or the control. Randomization helps us with our control. It helps us to eliminate the influence of other variables or the researchers and subjects.
The final thing that we're looking for is replication. If you complete the experiment several times, and you're able to use as many subjects as possible, you're getting a larger sample size. And you're able to be more accurate. All three of these things, control, randomization, and replication, help to have a well-designed experiment.
Let's go through an example to see how an experiment can be well designed. In this example, patients with Alzheimer's are divided into two groups. One group is given a new drug, the other a sugar pill. In order to exert control to eliminate the effect of other variables, we want to focus in very narrowly on the effects of the new drug and the old drug.
So we're going to group the participants into the same age range. This helps us to eliminate the effect of age so that it's not that the younger patients are responding better to the new drug versus the old drug or that there's more younger patients in the group of the new drug. By limiting the age range, we help to control the variables. We help to limit to look at just the old and the new drug. Similarly, if we thought that gender had an effect, we could help control the effect of gender by evenly dividing males and females between the two treatment groups.
As far as randomization, we would want to use a random number generator or double blind trials-- we'll talk about that in another tutorial-- in order to eliminate the chances of introducing bias. For example, if you knew one of the patients in the trial and really wanted them to do well, you'd want to make sure that they got the treatment.
This would affect the outcomes of the trial and would make it not as accurate because you'd be influencing it. You'd be exerting an effect, which is something researchers don't want to do. So randomization helps us to again control for some of the outside factors.
The final thing we'd want to do is make sure that our experiment is replicable. We'd want to use as many subject as possible. Sometimes, it's really hard to use a lot of subjects because of the costs involved. We'd also want to make sure that the treatments and that the experiment is replicable, that it's done more than one time, that it's duplicated.
And we can help to increase the number of our sample size and increase our accuracy by increasing either the number of people we give the original experiment to or the number of times the experiment is done. This has been your tutorial on experimental design.