I am having trouble conceptualizing normalizing versus controlling for variables and would appreciate some education on the matter.
For example, you want to do a regression with two variables:
- Hunger, which is how long you have gone without food.
- Working Memory, which is your score on a test of working memory.
You hypothesize that the amount of time you have gone without food will result in worse working memory, thus a negative relationship will be created: The more more hungry you are, the worse your working memory will be.
So, say you have reason to believe that Age, Sex, and Race all could be influencing both your working memory score and how hungry your participants are. You also have access to a database that can norm your working memory score, based on a large sample of healthy controls. What would you say are the correct methods for further analysis?
You run a regression on the raw WM scores and the raw Hunger scores, but you control for Age, Race, Sex.
You run a regression with age/sex/race normalized WM scores and the raw hunger scores, but do not control for any variables.
You run a regression with age/sex/race normalized WM scores and the raw hunger scores, but you still control for age, sex, and race.
I would greatly appreciate the community's wisdom on this matter and references where you can supply them, so I can better understand what is going on.