When you hear "Studies show..", do you think correlation or causation?

[HejKatt]

Example I heard on the radio this morning: A study of 1000 women showed that those who exercised at least 3x/week pre and post-partum lost weight faster, and were less likely to suffer post-partum depression.

 

I can understand the part about losing weight, but the (lower incidence of) PPD could also be explained by having friends or supportive family who could watch the kids so one could actually get out of the house and exercise. That is, correlation rather than causation.

 

Sometimes, I get tired of the stream of emails forwarded by well meaning friends to do X, eat Y because studies show Z. Do most studies show correlation at best? I can't seem to recall a study that showed causation. And if it's just correlation, I'm not sure I want to jump on the bandwagon and eat/buy/do the latest - I have enough on my plate (pun!) as it is. :tongue_smilie:

 

Does it show yet that I didn't take Statistics in school? :D Anyone want to chime in?

Edited October 5, 2011 by leeyeewah

[lionfamily1999]

After my eyes stop tumbling around in my head I tend to try and think of a few things that the study could also be showing - like you said, perhaps people who work out have more support, perhaps gyms breed friendship, &tc. I think that some studies are more effective than others, but for the most part they are ust inference that doesn't take into account all of the facts, just a small window. Studies show that washing your car will cause rain. ;)

[Renee in NC]

Almost always correlation. You have to control for other factors in order to have causation. It really depends on how well the study was designed and who the target population is.

[HejKatt]

After my eyes stop tumbling around in my head .. I think that some studies are more effective than others, but for the most part they are ust inference that doesn't take into account all of the facts, just a small window. Studies show that washing your car will cause rain. ;)

:lol: OK, glad I wasn't the only one rolling my eyes. At least, it's a good way to wake up in the morning!

 

Almost always correlation. You have to control for other factors in order to have causation. It really depends on how well the study was designed and who the target population is.

Yes, and that's really hard to tell from the soundbites we get on the radio/TV, isn't it? My problem is that i. I don't have time to read about every survey people tell me about, and ii. If I tried to read even one, my eyes glaze over rapidly (lack of Statistics knowledge, jargon).

[Rivka]

I never assume that a news report will give me enough information to figure it out. If I'm interested, or annoyed, I'll look for the actual journal abstract.

[LeslieAnneLevine]

It depends on how much I agree with what the studies show. :lol:

[Harriet Vane]

I think, "What study? How was it designed? Who participated? What controls were in place? How was the study conducted?"

 

Frankly, studies mean nothing to me until those questions are answered satisfactorily.

 

I tend to be the only freak in my acquaintance who fusses like that, though. :tongue_smilie:

[asta]

I never assume that a news report will give me enough information to figure it out. If I'm interested, or annoyed, I'll look for the actual journal abstract.

 

:iagree:

[Snickerdoodle]

When you hear "Studies show.."

 

When I hear that I think, "I can fit any scenario to fit my perception of reality."

[zenjenn]

"Studies" of this nature generally confirm what common sense would conclude, or are wrong.

[regentrude]

Most studies like this are not rigorous enough to show actual causality, only correlation. Popular medical information is notorious for this- partly because it is difficult to have enough study subjects (most studies are too small to be significant) and partly because controlling all aspects of a human subject is almost impossible.

[dirty ethel rackham]

I think, "What study? How was it designed? Who participated? What controls were in place? How was the study conducted?"

 

Frankly, studies mean nothing to me until those questions are answered satisfactorily.

 

I tend to be the only freak in my acquaintance who fusses like that, though. :tongue_smilie:

 

:iagree::iagree:

[HejKatt]

Most studies like this are not rigorous enough to show actual causality, only correlation. .

 

I think, "What study? How was it designed? Who participated? What controls were in place? How was the study conducted?"

 

 

I never assume that a news report will give me enough information to figure it out. If I'm interested, or annoyed, I'll look for the actual journal abstract.

 

Do you have favorite sources where the abstracts are linked, or studies are vetted for these criteria before being presented? Popular news doesn't seem to be coming off well here..

[KristinaBreece]

I always assume correlation, unless I have enough information on the issues and the outside factors to actually see proof of causation.

[AuntieM]

I think they often report correlation with a sensationalized angle trying to imply causation.

 

I don't take any study results seriously before knowing who conducted and/or funded it.

[Daisy]

If you want to know the truth, I usually think, "What stupid grant paid for THAT?"

[AuntieM]

:lol:

If you want to know the truth, I usually think, "What stupid grant paid for THAT?"

[WishboneDawn]

Almost always correlation. You have to control for other factors in order to have causation. It really depends on how well the study was designed and who the target population is.

 

Yup. Until I've had the opportunity to check a study out a bit more it's generally correlation for me.

[HejKatt]

If you want to know the truth, I usually think, "What stupid grant paid for THAT?"

 

:lol: I should have made a poll so you could answer Other!

[Harriet Vane]

:iagree::iagree:

 

Great minds think alike . . . :cheers2:

[lionfamily1999]

When I hear that I think, "I can fit any scenario to fit my perception of reality."

:iagree:

If you want to know the truth, I usually think, "What stupid grant paid for THAT?"

:lol:

[Paige]

I think, "What study? How was it designed? Who participated? What controls were in place? How was the study conducted?"

 

Frankly, studies mean nothing to me until those questions are answered satisfactorily.

 

I tend to be the only freak in my acquaintance who fusses like that, though. :tongue_smilie:

 

I agree. If a summary of a study interests me, I'll try to look up the details. If it is a well planned study, causation cannot be 100% proven but there can be strong evidence that the link is causal and not just a correlation. Sometimes, I'll conclude it is a weak study and it is likely just a correlation but other times I have been persuaded that there are causal relationships.

[PiCO]

Try watching this lecture on YouTube:

 

It's by the guy who did the FatHead movie. It has helped me a lot when I'm reading news articles about scientific studies.

[gardenmom5]

I can understand the part about losing weight, but the (lower incidence of) PPD could also be explained by having friends or supportive family who could watch the kids so one could actually get out of the house and exercise. That is, correlation rather than causation.

 

?

lower incidence of PPD among exercising women could also be higher levels of endorphins - so, causation.

 

1000 isn't a very big pool, and there's lots of stuff left out of their sample to make any findings.

[TravelingChris]

I do have lots of statistics and research methodology experience and I agree with most of the posters, correlation not causality. Doesn't mean that x doesn't cause y but the studies usually don't prove them. Now the giant studies- the big Boston woman's one (I believe about 100000) and the Danish cohort ones, I do think show or sometimes disprove causality. In particular, the Boston one showed the harms of hormone replacement therapy and the Danish one disproved the mercury/autism link.

 

Like Rivka, I check the original abstract, if I am interested in the findings.

 

I also tend to give more credence to meta analysis with lots of studies but only if the studies in the original were at least constructed fairly well. These are studies that use many small studies in the aggregate to come up with larger conclusions.

[Carpe]

I agree. If a summary of a study interests me, I'll try to look up the details. If it is a well planned study, causation cannot be 100% proven but there can be strong evidence that the link is causal and not just a correlation. Sometimes, I'll conclude it is a weak study and it is likely just a correlation but other times I have been persuaded that there are causal relationships.

 

This is basically how I feel.

 

In the case of exercise reducing PPD. If the study took 1000 women at risk for PPD and had them exercise 3x per week and they developed PPD less often than would be expected, then that would mean something. But just as likely they took 1000 random women and those who exercised 3x per week or more were less likely to get PPD, which is a meaningless result.

[texasmama]

I think correlation. Causation is much more difficult to "prove".

[Momof3littles]

I think correlation. Causation is much more difficult to "prove".

:iagree:

 

It is astounding how much garbage does get published even in "good" journals.