Not specifically low carb, but a really eye-opening discussion of dietary risk factors and how they are computed.

http://news.bbc.co.uk/2/hi/uk_news/magazine/7937382.stm

This is a good layperson explanation of relative versus absolute risks. Most epidemiologists and statisticians talk in terms of relative risks and hazards, thus the percentages and ratios. However, the numbers that are useful to a layperson are the absolute numbers. For instance, if something doubles my risk of a heart attack but I have an extremely low risk for a heart attack, then it might not mean much on an absolute scale. If my risk was 1 in 10,000 and now it's 2 in 10,000, that's not really anything to get concerned about, even though it sounds scary on a relative scale.

--Melissa

I like that. I get so irritated reading articles that are titled things like, "INTERVAL TRAINING INCREASES CALORIE BURN" and then goes on to say in a study of 500 people those who did interval training increased their weight loss by 5%" or something ridiculous like that. I can get a whopping 5% increase - really?!?! OH MY!!

Puh-lease. All this reminds me of a class on Statistics that I took in college. One segment that semester studied with a book called "Lying with Statistics." It basically showed you how you can take a statistic and twist it to say pretty much anything you want it to, and went on to show you example after example of just how that is done, every single day, especially in the mass media.

I used to use statistics a lot when arguing with people. Now I actively avoid them.

I agree. I think for most people who have no training in statistics (and I definitely include myself among that group!), it is easy to find yourself thinking: I have a 50/50 chance of getting a terrible disease, if I double that risk I will be 100% guaranteed to get it! We might not put it in those words, but that is the underlying, unexamined feeling that gives these statistics such power among the numerically illiterate.

When they did epidemiological studies on cigarettes they found staggering numbers, like smoking increased cancer rates by %1000 percent (can't remember exactly but the number was huge). Now that's useful.

I hear that I have a close to 100% chance of dying someday!

Yes. It is easy for me to parse these media reports for what they are--I'm a statistician with training in epidemiology. Unfortunately these reports are just designed to grab attention and sell newspapers/tv. Sensationalism sells, even if it's misleading or not particularly informative to laypeople. I stopped watching the local news years ago. I got sick of seeing headlines for "news" on things like Nose Hair: The Hidden Danger!



Yes, the numbers for smoking were massive. A cursory search pulled up incidence rates from a recent prospective cohort study. The study separated men and women to see if there was a difference, but the differences between smokers and non-smokers are staggering. An excerpt from the abstract:

http://www.ncbi.nlm.nih.gov/pubmed/18556244



So lung cancer in non-smokers occured at a rate of 20-25 per 100,000 person years vs. 1259-1309 per 100,000 person years for smokers. That is 52 TIMES the risk, which is not in the same league or even playing the same game as an increase of 20% in risk.

Epidemiology methods work very well for detecting infectious diseases, which was their original purpose. They also work well when the risks are obvious, like for smoking, arsenic exposure, etc. Where they are less useful is with chronic diseases where the increased risk may be subtle and significant only in the context of an entire population. The problem is that these are the ones that we hear about in the media these days.

The media conveys the results of studies in such a way to imply that we should immediately change our behavior. This is an awful idea--one should never make health decisions based on one study. As we have seen from some of the many articles passing through this sub-forum, any given study can be statistically impeccable or a shoddy example of science. And we should definitely not be making policy decisions this way either. There have been studies in the past that showed the opposite relationship from what was later found to be the case. The Ancel Keys Seven Countries study is a great example of this. Fool me once, shame on you--fool me twice...

--Melissa

It's more like "Fool me twice a day, business as usual".

I love what Gary Taubes wrote about epidemiology and how we're constantly misusing it: http://www.nytimes.com/2007/09/16/m...emiology-t.html

Yeah it's a good article. Epidemiology is great for its intended purpose, but its intended purpose was never to produce studies that lead to policy decisions or health decisions. Epidemiologic studies are supposed to be hypothesis-generating--then you do randomized controlled trials to test these hypotheses. And you should do multiple RCTs before making any policy or health decisions based on the results. People often try to shortcut this process when there is no suitable shortcut. Nothing worth doing is easy.

--Melissa

It can be very dangerous as well. Most pregnancy complications are presented prefaced by the words, "over weight and obese women a x amount more likely to have such problem", even when the risk difference is very small. I can't tell you how many people I've spoken to or read on pregnancy forums who thought they were immune because they were thin from things like hypertension or gestational diabetes.

Agreed. Just because obesity makes x more likely does not mean that thin women can't get it too. Non-smokers do get lung cancer--just look at Christopher Reeves's wife.

--Melissa

Non-smokers get COPD too, like my Mom.

Thought that this interesting article probably belonged here:


Science journalists? Don't make me laugh

Ben Goldacre

Science is not difficult to explain. Today we will see how British journalists go out of their way to cherry-pick which evidence they cover, and then explain the risks and benefits in what has been shown to be the single most unhelpful way possible.

The article continues here: http://www.guardian.co.uk/commentis...prostate-cancer

The book "Seeing Through Statistics" by Dr. Jessica Utts is a good read for the layman. She is really good, in person as well, at describing statistics in a way even total layman can understand. As statisticians go there are two primary kinds: mathematical sorts, and conceptual sorts (to whom the math is only a tool, not the primary study). She is the latter.

I think people learn more about "intelligent discretion and logic" in that kind of statistics than about math, since it's really more about how to evaluate reality in-context.

Statistics used to be part of the occult, because of its predictive abilities. I still find that funny.

PJ