We don't want confounding variables to affect the response variable in an ideal experiment, yet we know that sometimes they are going to exist anyway. How can we set up the experiment so that the effects of confounding variables are minimized? There are three ways...
1. Randomization. When we randomly place subjects into treatment groups, we are mixing up the members of our sample so that they are diverse.
2. Replication. This means performing an experiment many times. Ideally, when we replicate an experiment over and over, we are looking for similar results each time. What if we don't see similar results? Then there may be a confounding variable within our data, or perhaps the experiment needs to be re-designed.
*3. Blocking. Once you identify a possible confounding variable (i.e. gender) we can create experimental groups based on that variable (i.e. one group for boys and one for girls). Then we randomize within those groups. Perform the experiment, then compare your groups at the end to see if there are differences between boys and girls.
Note: The blocking variable does not have to be gender, I used that as an example because I thought it would be easy for everyone to understand. Your blocking variable should be something that could potentially confound your study.
No comments:
Post a Comment