John, a medical practitioner at a private medical facility, is discussing the problem that one of his patients, Tom, says he has been having with the medication that the doctor has prescribed. It is a new medication and the facility has been asked to gather information about the medication for the manufacturer. At a previous visit to the doctor, Tom was complaining that the medication made his stomach upset. A check of the medication reactions data shows 0% reports on stomach upset in extensive testing. The doctor has asked Tom to fill out a log sheet of when he takes the medication and when, after that, that he feels his stomach get upset.
John takes the log sheet and gives the data to his doctor. The agency he sends it to calculates a correlation coefficient for the data. The calculation shows that there is a positive 0.85 correlation between taking the medication and stomach upset.
Tom sees the number and says, "See, that proves it. The medication causes stomach upset!"
What do you have to say about Tom's statement?
The issue here is the relationship between correlation and causation. In this case, although we have a large (and presumably significant) correlation between taking the medication and stomach upset, we haven't had enough control over the situation to conclude that the medication CAUSES stomach upset, as Tom states. There are several other variables that ...
Explains the issues relating to causal conclusions using an example of a correlational design.