Correlation to causation?

[basophilic]

In what cases can a correlation become considered a causation? Also, in general are most natural science experiments causation-based and social science ones correlation?

---

Members don't see this ad.

 

[desertrat12]

Do you have a specific example where this question came up? What is your motivation behind this question? I'm trying to see if you encountered a particular question/example/passage where something was said to be causation, and thus your question about causation was generated. I ask this because unless there is something specifically telling you (like a passage quote) that some relationship is causation, you shouldn't interpret a scientists data (natural science or social science) as being causative. To say something is causative is a very large jump to make and it ignores the possibility of a confounding variable, and involves a bunch of other problems as well. Generally speaking science is never 100% conclusive.

 

[basophilic]

Do you have a specific example where this question came up? What is your motivation behind this question? I'm trying to see if you encountered a particular question/example/passage where something was said to be causation, and thus your question about causation was generated. I ask this because unless there is something specifically telling you (like a passage quote) that some relationship is causation, you shouldn't interpret a scientists data (natural science or social science) as being causative. To say something is causative is a very large jump to make and it ignores the possibility of a confounding variable, and involves a bunch of other problems as well. Generally speaking science is never 100% conclusive.

Don't have specific quotes/questions, but wondering in general. So technically nothing is entirely causative? Also, I'm guessing physical sciences have highest level of causation, lesser causation in bio sciences, and least in social sciences.
And to truly prove causality, you have to do as many experiments as needed to control for every single confounding variable?
Also, to prove causality A --> B (for any physical/bio/social area):
1. You have to do large number of experiments to control as many confounders as possible.
2. Each experiment must demonstrate high internal validity by doing things like choosing a highly homogenous sample and then completely randomizing samples to the control and experimental groups.
Is that fair?
I know these are a bit random topics and prob obvious to most, but just trying to get a better perspective.

 

[desertrat12]

You are definitely asking good questions, especially the way you are tying in different principles to help get at a complete answer (your other posts demonstrate this as well), this is exactly was is needed for the mcat, good application and synthesis of lots of info, so nice work. As for your question in this post, someone much more qualified in research than myself can give you a better answer, but from my understanding it is usually a safe bet to stay away from the idea that a relationship is causative. The questions that I have seen will often go as follows: based on figure 2 what conclusion is most reasonable? Often the answer choices will have one that claims causation and one that claims some type of correlation. I've never, never, seen it where causation was the correct answer. It makes for a great distraction though. There definitely is the possibility of exception, but I'm not well versed enough in research to give you an example. Staying away from causation was sufficient for me while studying/taking the mcat.

 

[Labrat07]

I learned something on this. Thank you