One of the most commonly used tactics by Aubrey and Michael, and Maintenance Phase fans in the MP subreddit, is to dismiss a scientific finding by saying “correlation =/= causation” or saying that all of the existing studies are just correlations. In fact, this saying has become so commonplace on the internet that Slate wrote a
I love how I’m not just learning exactly what Maintenance Phase gets wrong in its statistical “analysis” but also I am learning so much about statistics through this little series. Thank you!!!!
Thank you so much for commenting, Elizabeth! It makes me so happy to know that you are finding this helpful! I'm hoping to post some more of these short ones just to provide some more resources for people. :)
Holy geez: I really hope that your actual professional life is centred around medical/scientific communication, bc boy do ever have a gift for it.
Nothing to add (dated an epi prof for years, so have spent enough of my life teasing out the nuances of statistical methodology!), but just wanted to drop a note of thanks for taking the time to put this kind of info out into the ether in such a thorough yet easily digestible format.
Holy geez this is one of the best compliments I have ever received! Thank you for reading and thank you for saying that. I am really passionate about this stuff and I don't really get to do it enough in my day job so it's fun to have this outlet!
First of all, I don't sense that you're commenting in good faith, because one cannot in good faith argue that it "makes perfect sense" to refer to correlations when correlations weren't used, so I'm hesitant to engage with you. But in the interest of keeping an open mind, I will respond to your comments. It is clear that you're not an epidemiologist or biostatistician. Causal inference is a robust field of methodological research that, in part, attempts to address the presence of residual confounding. Importantly, residual confounding does not necessarily "render such inferences dubious." It varies by study and strength of the association, etc. By your logic, only double-blind placebo-controlled randomized clinical trials would produce valid evidence. Do you think that smoking does not cause lung cancer because we didn't do an RCT to find that out? I'd hazard a guess the answer is no. Further, I'm not sure if you caught the fact that I have a PhD in this stuff, so I don't need a blog post from a controversial influencer economist to explain confounding to me. She's fundamentally wrong that observational data can't provide a causal link. Please look into Miguel Hernan's work or any of the references I provided in the post above.
I'm sorry that you don't like my post! I'm not sure which article you mean by "the first article you reference". Do you mean the Slate article? Or the one by Fedra Negri? I cited both of those purely as examples of how ubiquitous this phrase has become. That's all. You don't have to think the article is good for it to be an example of academics and media addressing the fact that people overuse this statement. That being said, I don't see anywhere in either article where it suggests that there cannot be causation without correlation. Can you point me to where you see that?
I think you may have misunderstood the point of this post - I didn't even try to explain how correlation can imply causation, nor did I indicate that I would be doing that. The whole point of this post is to explain why the phrase "correlation =/= causation" is not always applicable and highlight that it is often misused. Anyway, the thing about explaining things is that not every explanation works for every person! It's clear from the other responses that I'm getting that a lot of people understood my post and appreciated it. You don't have to read if you're not getting anything out of it!
I think we'll have to agree to disagree that economists are at the cutting edge of causal inference. ;) Econometrics methods are quite different from epidemiologic methods. For example, while IVs are really popular in econometrics, they are pretty controversial and very rarely used in the epi literature. I think you're coming from a different background. Which is fine! But that's not what this post is about. You have no obligation to read my content. Many people have told me that this is helpful to them and they appreciate my explanation. :)
I'm not sure what it is that you disagree with. Nothing that I said is really an opinion - I'm just explaining what correlation really means and why saying "correlation =/= causation" is not always relevant. It has become a pithy thing to throw out when people disagree with research findings, and I think most people don't really know what it truly means. The point of the post was to educate folks. The fact that laypeople use "correlation" interchangeably with "association" is exactly the point that my post is intended to address.
Hernan's entire book is about causal inference, so I'm not sure what you mean that you didn't think the chapters you read were relevant to this question. You have very strong opinions about these things for someone without training or expertise in the field and I'm not sure what continuing this discourse could possibly achieve, so I'm going to bow out. If you're really interested in learning more about causal inference, you can look into all the references I've provided already!
I am not focused on your lack of credentials - I'm saying it is difficult to address your questions because they are based on uninformed opinions (as you'll see, I gathered that based on your original comment, prior to even knowing your "credentials" and taking a graduate level course does not make you an expert). I don't know how to address your points because they don't apply to what I wrote. I'm not asserting anything here that isn't already consensus among experts (see all of the references I provided). I value constructive feedback - you can see that in my responses to other commenters. But your comments are a strange form of concern trolling and I simply don't know what you want to get out of me. Sorry for coming across as unkind - I just really don't get what you are angling for.
I love how I’m not just learning exactly what Maintenance Phase gets wrong in its statistical “analysis” but also I am learning so much about statistics through this little series. Thank you!!!!
Thank you so much for commenting, Elizabeth! It makes me so happy to know that you are finding this helpful! I'm hoping to post some more of these short ones just to provide some more resources for people. :)
Holy geez: I really hope that your actual professional life is centred around medical/scientific communication, bc boy do ever have a gift for it.
Nothing to add (dated an epi prof for years, so have spent enough of my life teasing out the nuances of statistical methodology!), but just wanted to drop a note of thanks for taking the time to put this kind of info out into the ether in such a thorough yet easily digestible format.
Holy geez this is one of the best compliments I have ever received! Thank you for reading and thank you for saying that. I am really passionate about this stuff and I don't really get to do it enough in my day job so it's fun to have this outlet!
First of all, I don't sense that you're commenting in good faith, because one cannot in good faith argue that it "makes perfect sense" to refer to correlations when correlations weren't used, so I'm hesitant to engage with you. But in the interest of keeping an open mind, I will respond to your comments. It is clear that you're not an epidemiologist or biostatistician. Causal inference is a robust field of methodological research that, in part, attempts to address the presence of residual confounding. Importantly, residual confounding does not necessarily "render such inferences dubious." It varies by study and strength of the association, etc. By your logic, only double-blind placebo-controlled randomized clinical trials would produce valid evidence. Do you think that smoking does not cause lung cancer because we didn't do an RCT to find that out? I'd hazard a guess the answer is no. Further, I'm not sure if you caught the fact that I have a PhD in this stuff, so I don't need a blog post from a controversial influencer economist to explain confounding to me. She's fundamentally wrong that observational data can't provide a causal link. Please look into Miguel Hernan's work or any of the references I provided in the post above.
I'm sorry that you don't like my post! I'm not sure which article you mean by "the first article you reference". Do you mean the Slate article? Or the one by Fedra Negri? I cited both of those purely as examples of how ubiquitous this phrase has become. That's all. You don't have to think the article is good for it to be an example of academics and media addressing the fact that people overuse this statement. That being said, I don't see anywhere in either article where it suggests that there cannot be causation without correlation. Can you point me to where you see that?
I think you may have misunderstood the point of this post - I didn't even try to explain how correlation can imply causation, nor did I indicate that I would be doing that. The whole point of this post is to explain why the phrase "correlation =/= causation" is not always applicable and highlight that it is often misused. Anyway, the thing about explaining things is that not every explanation works for every person! It's clear from the other responses that I'm getting that a lot of people understood my post and appreciated it. You don't have to read if you're not getting anything out of it!
I think we'll have to agree to disagree that economists are at the cutting edge of causal inference. ;) Econometrics methods are quite different from epidemiologic methods. For example, while IVs are really popular in econometrics, they are pretty controversial and very rarely used in the epi literature. I think you're coming from a different background. Which is fine! But that's not what this post is about. You have no obligation to read my content. Many people have told me that this is helpful to them and they appreciate my explanation. :)
I'm not sure what it is that you disagree with. Nothing that I said is really an opinion - I'm just explaining what correlation really means and why saying "correlation =/= causation" is not always relevant. It has become a pithy thing to throw out when people disagree with research findings, and I think most people don't really know what it truly means. The point of the post was to educate folks. The fact that laypeople use "correlation" interchangeably with "association" is exactly the point that my post is intended to address.
Hernan's entire book is about causal inference, so I'm not sure what you mean that you didn't think the chapters you read were relevant to this question. You have very strong opinions about these things for someone without training or expertise in the field and I'm not sure what continuing this discourse could possibly achieve, so I'm going to bow out. If you're really interested in learning more about causal inference, you can look into all the references I've provided already!
I am not focused on your lack of credentials - I'm saying it is difficult to address your questions because they are based on uninformed opinions (as you'll see, I gathered that based on your original comment, prior to even knowing your "credentials" and taking a graduate level course does not make you an expert). I don't know how to address your points because they don't apply to what I wrote. I'm not asserting anything here that isn't already consensus among experts (see all of the references I provided). I value constructive feedback - you can see that in my responses to other commenters. But your comments are a strange form of concern trolling and I simply don't know what you want to get out of me. Sorry for coming across as unkind - I just really don't get what you are angling for.