causation vs correlation

[kristinaroz93]

I find this topic to be confusing. The lsat does a good job with tricking people in this regard. So how can I really tell the difference between these, do I need the actual terms "caused" to infer causation, or what else do I look for when looking to see that something is actually causation and not just a correlation.

1)For example, upon doing a practice test I saw this in the lr section "The more cholesterol we have in our blood, the higher the risk we shall die of a heart attack". It is really tempting to infer causation had the phrase "the higher the risk" not been supplied. Since it reads if a happens then b happens. Would this be a causation situation in that case:"if we have more cholesterol in our blood, then we shall die of a heart attack"? Or do we physically need the term "caused" to infer causation and that without it we have a correlation regardless even if taking the phrase "the higher the risk" out (in other words taking out the probability from the sentence and making it more assertive).

2)And also going back to this problem since we are on the topic (which is from the october 1999 lsat -"these days, drug companies and health professionals alike are focusing.."), we eliminate A because causation/correlation aside there are many other things that lead to heart problems, so even in the event high cholesterol does cause heart disease, if you do not have high cholesterol it does not mean your risk is low as there could be other things that contribute to the same problem? How is my reasoning here?

3) And E is right, since smoking, drinking, and exercise (i.e. lifestyle choices) influence cholesterol which increases risk of heart problems, so altering any of those lifestyle choices can affect our probability of getting or not getting heart disease via its effect on cholesterol. Just making sure here as well!

(I know I am over analyzing on these easy problems, but I think they can shed light on bigger things that can help with harder problems).
This topic is still confusing to me and I would like to truly understand how to properly infer whether something is causation and or correlation, and how the problem should be approached differently upon making that inference. Thanks in advance!

[Nikki Siclunov]

Hi Kristina,

Thanks for your question! Yes, correlations and causality are frequently encountered on the LSAT, and it's easy to confuse the two. Before I address your specific questions, let's get a few things straight:

1. In statistics, "correlation" refers to a relationship between two interdependent variables (e.g. height and weight, studying and grades, etc.). A correlation alone does not prove a causal relationship, but it can suggest that a causal relationship does, in fact, exist. For instance, a correlation between height and weight cannot be interpreted to mean that either one causes the other. On the other hand, a correlation between studying and grades implies that studying causes higher grades, but the correlation alone is not sufficient to prove that such a causal relationship exists.

You can think of it that way: an empirically observable correlation between two interdependent variables is a necessary, but not sufficient, condition for causality.

2. By contrast, causation is usually (though not always) a conclusionary idea: we conclude that studying causes higher grades, because we know that the two are correlated. Conveying a causal relationship typically involves the classic cause/effect indicators (due to, caused by, responsible for, etc.), but again - there are exceptions to this. Oftentimes, a causal relationship is assumed to exist, but is not overtly stated as such. Let me give you an example:

"Those who study more than 20 hrs/week tend to receive higher grades than those who do not. Therefore, I should probably study at least 20 hrs/week"

This argument is actually causal, even though none of the classic causality indicators are present. Why? Because the conclusion clearly assumes that studying causes one to receive higher grades. Without such an assumption of causality, the conclusion wouldn't make any sense.

Bottom line is this: Correlations are not questionable, as they are usually presented as empirically observable evidence. By contrast, causality is questionable, in so far as a causal conclusion is difficult to formally deduct from the evidence presented. Consequently, the more definitive the causal conclusion, the weaker the argument is. The more probabilistic or inductive that conclusion is, the stronger it is.

Now, let me address the specific example you mention:

The stimulus contains a fact set. Strictly speaking, the following statement of fact conveys a correlation, not a causation:

The more cholesterol we have in our blood, the higher the risk we shall die of a heart attack

Is high cholesterol responsible for elevating the risk of heart attack? Possibly. But it's not something we can absolutely prove. All we know is that the two are correlated.

Now, take a look at the last sentence in the stimulus: "At least three factors - smoking, drinking, and exercise - can influence the levels of cholesterol in the blood." Now, this fact is clearly causal. The stimulus does not contain a conclusion, and so we are required to assume that all the statements in it are actually true. Consequently, we must assume that the following causal relationship is true:

smoking/drinking/exercise (cause) cholesterol levels (effect)

So, what sort of a conclusion can we draw from these two facts? If cholesterol levels correlate with the risk of heart disease, and certain lifestyle factors can influence cholesterol levels, then it is reasonable to conclude that these factors could also influence your risk of heart disease. This conclusion is justified not because high cholesterol necessarily causes heart disease, but because the two are correlated. This validates answer choice (E).

Answer choice (A) is incorrect, but for a slightly different reason than the one you state. The stimulus indicates that "the more cholesterol you have in your blood, the higher the risk you will die of a heart attack." So, can we conclude that the lower your cholesterol levels, the lower your risk? Possibly. Of course, there could be other factors affecting your risk of a fatal heart attack, but it is beyond doubt that the two are, to an extent, correlated. So, why is answer choice (A) not necessarily true? Because the correlation is presented in comparative, not absolute terms! Even if lower cholesterol correlates with a lower risk of a fatal heart attack, it is still possible that having low cholesterol does not guarantee a low risk of a fatal heart attack. It is entirely possible that below a certain threshold, the two are no longer in sync. Also, notice how answer choice (A) is presented as a conditional statement: low cholesterol low risk. The thing is, such an absolute relationship cannot be deduced with certainty given the information presented. Low cholesterol is neither a sufficient, nor a necessary condition for a low risk of heart disease.

One final remark here: Must Be True questions where the stimulus contains a fact set invite us to put the facts together and formulate a reasonably drawn conclusion based on them. Definitive statements are a lot more difficult to prove than probabilistic ones! Take a look at the language in answer choice (A) vs. (E). In answer choice (A), the author states an absolute relationship in no uncertain terms. Can we deduce it with certainty? Nope. By contrast, answer choice (E) uses the word "can": the risk of a fatal heart attack can be altered by certain lifestyle changes. Of course it can! This is a matter of possibility, not absolute certainty. If answer choice (E) had said something along the lines of, "Those who don't smoke or drink, but instead work out 24/7, must have a very low risk of heart disease" - do you think we can really prove that? I would guess not.

Check out these blogs on the topic of correlations and causation:

http://blog.powerscore.com/lsat/bid/212 ... SAT-Part-I
http://blog.powerscore.com/lsat/bid/220 ... AT-Part-II
http://blog.powerscore.com/lsat/bid/229 ... T-Part-III

Hope this clears things up a bit!

[kristinaroz93]

Hi Nikki,

Your explanation here was incredible! I thank you for the time you took to write everything and explain it as you did, it really helped me! I have a better feel for the difference between those two factors now. And as for A I understand what you are saying. While the reason I gave had partially to do with it, it was not the best nor strongest reasont to eliminate A. The best reason was that A was too strong/definitive based on the information in the stimulus. From the statement, "The more cholesterol we have in our blood, the higher the risk that we shall die of a heart attack", there is not way to conclude based on this correlation that "low blood cholesterol means your risk of fatal heart attack is low". It could certainly be lower risk, but not necessarily "low risk". I hope I understood you well in regards to choice A, let me know=)

Thanks Again!

[Nikki Siclunov]

Kristina,

That's exactly what I meant Glad this helped... causality is such an important paradigm in Logical Reasoning that it really pays to understand it.

Best of luck!