ExplanationChoosing Your Experimental Design
Before you start an experiment, it is a good idea to put some thought into which type of experimental design you will use. Both have advantages and disadvantages and sometimes you are forced to choose one over the other. Here are the usual scenarios in which the choice of experimental design is easy:
There are some other considerations when choosing an experimental design. The main ones are:
- When the dependent variable cannot be manipulated, you probably need an independent design
A good example of this point would be an experiment that compares males with females. You cannot test the same people under each condition (male and female) as you cannot change their gender, so unless you are pairing them in some other way (twins, for example) you must compare one group of males with a group of different females.
- When you cannot repeat a measurement on the same experimental unit, you probably need an independent design
When using a paired design, you often test (or measure) the same experimental units twice. If the first test is likely to have an effect on the second test or if the unit can be tested only once, then the second, repeated test is not possible. Crash testing cars would fall into this category as you cannot crash test the very same car under two different conditions - the first crash rules out the second test!
- If you need to test the same experimental unit under different conditions, you need a paired design
- It is easier to control for confounding variables with a paired design. Anything other than the independent variable that might affect your measurements is referred to as a confounding variable. There is always a risk that an experiment might produce a difference that is due to something other than the independent variable. For example, if you wanted to compare the heights of males and females but you chose one group from a basketball team and the other from a group of jockeys, you might find a difference, but would it be due to gender? By measuring the same subjects twice, you reduce the risk of introducing confounding variables.
- If subjects are hard to find, then testing the same ones twice obviously doubles the quantity of data you can collect. It has the disadvantage of requiring each subject to return to produce the second sample, which can increase the risk of you having to throw away any data from the first sample that does not have a paired value in the second.
- There may be ethical issues which either prevent you from manipulating the independent variable or dictate that subjects must be tested under both conditions.
- If you are testing people, you may find that giving them the same test twice under different conditions allows them to practice the task in the first condition, which naturally produces an improvement in the second condition. Similarly, subjects might grow bored of the task and show fatigue effects when asked to perform it under the second condition.
Here are the advantages and disadvantages of each type of experimental design:
|Within Subjects||Between Subjects|
- Fewer subjects are needed as each subject is tested twice;
- You have more control over confounding variables.
- There is less risk of practice or fatigue effects;
- There is less risk of data loss due to drop-out as subjects are only measured once.
- Subjects may drop out, not completing the second condition and so rendering the data from their first condition unusable;
- Subjects can suffer from practice or fatigue effects when tested twice.
- Twice as many subjects are required;
- You have less control over confounding variables.