Describe some of the different sources of confounding variables in both between-subjects and within-subjects designs.
How can researchers work around these potential pitfalls when conducting an experiment?
What about when using single subject designs?
Finally, when would you suggest using a between-subjects experimental approach instead of a within subjects one, and what about vice versa?
Let's start by looking at how each design works. Imagine, for example, that you have developed a new 'quit smoking' program. You want to know if it's better than an existing, older 'quit smoking' program. To test this, you'll select people who are currently smokers, assign them to a treatment (program), and then count the number of people who are still smoking, say, 6 months later.
In a between-subjects design, you'll have half of your smokers assigned to the new program, and the other half assigned to the old program. You'll compare smoking rates between the two groups of DIFFERENT people after 6 months of their respective treatment. The idea is that any differences between the groups must be due to the differences in treatment.
In a within-subjects design, all of your smokers would complete each program, in sequence. The best way to do this would be to counterbalance the order of treatments (half of them do the new program followed by the old, and the other half do the old followed by the new). Again, differences in smoking rates after each type of treatment would be attributed to the effects of the quitting programs.
In between-subjects designs, we run into a key confound:
We are testing different people in each group. We are assuming that our groups are equivalent, and this is probably basically true as long as we've randomly assigned people to each group (this is one way ...
Explains the advantages and disadvantages of within-subjects and between-subjects designs. Also addresses single-subject (ABAB) designs.