I have seen this little catch phrase thrown around in way too many situation where it is not appropriate. I have concluded that most people have an idea what it means, but have not fully grasped the concept. Let me explain.
In order for a correlation to exist between two variables, you first need a set of data. And more than 1 row of data. Only then can you begin to look for a correlation. And *if* you even have such data to suggest a correlation between two variables, you can finally start to think about whether there is causation.
Many people like to feel smart by saying "correlation does not equal causation", but really there is not even any correlation present in the first place. Like in the n=1 case studies that are so common throughout the forum, there is just not enough data to suggest a correlation. So in those instances, when you say "I know, correlation is not the same as causation", you are just demonstrating that you do not understand the concept. Because you are implying that there IS a correlation, when there is not. And if there is no correlation, then there can be no causation.
Here is an instance when you need to start thinking about whether a correlation implies some sort of causation. Assume we had the following (fictitious) data:
60% of people with shitbrain disease die from lung cancer.
There is clearly a correlation here. However, some people might assume that shitbrain disease causes lung cancer and death. BUT that is not necessarily true. Later on, we find some more data:
90% of people with shitbrain disease have an unexplained urge to smoke cigarettes, and begin smoking 1 pack per day.
The likely conclusion is that smoking causes lung cancer and death. Shitbrain disease causes people to smoke, and smoking causes people to get lung cancer and die. There is a clear correlation between shitbrain disease and lung cancer, but shitbrain disease does not cause lung cancer.
Now, if our data showed that only one person that has shitbrain disease got lung cancer, that still does not suggest even the weakest correlation between shitbrain disease and lung cancer. So in that case, if you say "correlation does not equal causation" you would be making yourself sound silly.
So please, don't make yourself sound stupid by saying that "correlation does not equal causation" in situations where there is not even a weak correlation.
In order for a correlation to exist between two variables, you first need a set of data. And more than 1 row of data. Only then can you begin to look for a correlation. And *if* you even have such data to suggest a correlation between two variables, you can finally start to think about whether there is causation.
Many people like to feel smart by saying "correlation does not equal causation", but really there is not even any correlation present in the first place. Like in the n=1 case studies that are so common throughout the forum, there is just not enough data to suggest a correlation. So in those instances, when you say "I know, correlation is not the same as causation", you are just demonstrating that you do not understand the concept. Because you are implying that there IS a correlation, when there is not. And if there is no correlation, then there can be no causation.
Here is an instance when you need to start thinking about whether a correlation implies some sort of causation. Assume we had the following (fictitious) data:
60% of people with shitbrain disease die from lung cancer.
There is clearly a correlation here. However, some people might assume that shitbrain disease causes lung cancer and death. BUT that is not necessarily true. Later on, we find some more data:
90% of people with shitbrain disease have an unexplained urge to smoke cigarettes, and begin smoking 1 pack per day.
The likely conclusion is that smoking causes lung cancer and death. Shitbrain disease causes people to smoke, and smoking causes people to get lung cancer and die. There is a clear correlation between shitbrain disease and lung cancer, but shitbrain disease does not cause lung cancer.
Now, if our data showed that only one person that has shitbrain disease got lung cancer, that still does not suggest even the weakest correlation between shitbrain disease and lung cancer. So in that case, if you say "correlation does not equal causation" you would be making yourself sound silly.
So please, don't make yourself sound stupid by saying that "correlation does not equal causation" in situations where there is not even a weak correlation.
