Where we get fit and spin (wool)

Posts tagged ‘causation’

Statistics and Lies

Mark Twain did not care much for statistics, he said:

Get your facts first, then you can distort them as you please.


There are lies, damned lies and statistics.

I have reason to agree, when simple curiosity drove me to do some quick research. Every time there is an article on health, I read it. The one that caught my eye today was one on Yahoo that another country beat out the United States as the fattest country. The main thrust of the article was that Mexico now has a greater proportion of obese people than the U.S. .This made me think about the link between obesity and heart disease, wondering where Mexico and the United States fall in the ratings of heart disease, so I went looking for statistics on heart disease to see what sort of correlation there was. That is where things got interesting. I googled it and clicked on the first two links. Both claimed to use the latest sources, both cited WHO, the World Health Organization, and yet they don’t have the same information. Part of this could be what filters they use, and that is part of the problem with statistics, they are endlessly maniple. The first site I went to worldlifeexpentancy.com, listed Turkmenistan as the number one country for coronary heart deaths, and Slovakia was way down on the list. The second site, nationmaster.com, listed Slovakia first, and Turkmenistan wasn’t even on their list. The initial article said that Japan was the slimmest country, and it was way down on both lists for heart disease, but not at the same point, and the number of deaths per million differed on the two lists.

I went to WHO’s website to try to find their original source, but either they don’t put it up there, or you can’t find it with a search engine. The closest I could get was a listing of CVD and diabetes. If anyone has a link to it, please put it in the comments, I would be interested.

Getting back to my original article, it raises some interesting points about the interpretation of facts:

According to a new report from the United Nations Food and Agricultural Organization, the United States is no longer the world’s fattest developed nation―Mexico is.

The key word here is developed. Take that out, and you go from Mexico’s 32 % to :

The world’s fattest nation overall is Nauru, a South Pacific island where a staggering 71.1 percent of its 10,000 inhabitants are obese.

The U.N. report does not include data for American Samoa, which has been tabbed in the past as the world’s fattest country. According to a 2010 World Health Organization report, nearly all of that Pacific island’s inhabitants (95 percent) are considered overweight.

So one world makes a word of difference in what we are talking about. To quote Mark Twain again, “it’s the difference between lightning and a lightning bug”. There are many points to be made from this little exercise.

  • People are doing their best to study health and fitness, the shifting recommendations come from a variety of factors. A lot of smart people are spending a lot of time gathering data, and trying to come up with study designs that compensate for the flaws of previous ones.
  • After we’ve collected the data, It has to be analyzed, and that is fraught with hazards. That is where bias can wreak havoc.
  • Every word can count. What factors are included/excluded from the study?
  • News reporters just read the summaries, they don’t sift through the data to see if the summary is justified. A study might show a weak correlation between two things, and the news article may read “A causes B!”
  • Heart disease is affected by a lot more than just weight. The top countries for heart disease have high rates of smoking, and less health care.
  • The furthur you get from primary sources, the more corrupted the data.

It is hard to when the information changes over time, and it will. I don’t want anyone to get discouraged, just don’t put all your faith in one study, or get cynical when a study gets discredited.  It takes a whole lot of research to get any kind of picture of what is really going on.

Many of you may have heard about the link between omega 3’s and prostate cancer. The whole fish oil craze started with studying people who ate high amounts of cold water fish, and their lower risk of heart disease. They then took the fish into the laboratory to see what aspect of it was causing this. Once the oil was isolated, we then started to put it in a pill, for those who don’t like fish. We can’t study every aspect at once, so the prostate link wasn’t found right away. We are starting to see that even if there is one active compound in a food, it doesn’t seem to have the same effect once isolated that it appeared to have in the food.

The last point I’d like to make on this is that we will all die of something, at some point. All we can do is try to improve the quality of life until the end. No food will extend your life to extreme levels, and there are a variety of factors that influence health and life expectancy. Pills are not magic, and they all have side effects. You are far better off eating the best you can and not worrying about the rest. By the way, if you read that  link to the 2010 WHO report inside the quote I put in this article, it will dispell any doubts you may have about the link between diet and life expectancy.

Correlation does not Equal Causation

This article from the wall street journal might be misleading. I can’t say it is, since I have not seen the study. It supposedly shows a link between increased red meat consumption and the increase in type II diabetes. Researchers found a link between increased meat consumption, by as little as 1/2 a serving a day, and a greater risk of diabetes.

My question is, did they factor in the greater calories and weight gain  that increasing your consumption of any food would cause, since we know that being overweight will lead to greater diabetes risk? Did they separate out the consumption of “real” red meat vs processed, like hot dogs and lunch meat? It stated it was independent of body weight and quality of the rest of the diet, but what does that mean? It is also based on questionnaires, which are notoriously inaccurate.

The trouble with reading an article about a study is that you are getting conclusions twice removed. First, the authors of the study write a conclusion about the facts in their study. They can be drawing incorrect conclusions, or at least be suffering from bias. Then you are getting the conclusions that the author of the article drew from the conclusions of the authors of the study. Without being able to see the original study and taking the laborious amounts of time to go through the data to see if their conclusions are correct, you have to take some of this with a grain of salt. I have read enough arguments over studies to know that “facts” can be created from poor data collection, bias, poor study set up, and more. At least the author of this article did furnish a comment from a detractor of the study, so you have an idea of what other opinions are out there.

This is not to say that I am championing eating red meat in large quantities. Enough studies have been done to prove that less is more when it comes to red meat. However, it does seem counter intuitive to link red meat to diabetes, since it is sugar regulation that is affected by diabetes. I would love to know if they did control for the calorie increase, and how much the participants weighed, since being overweight is a strong factor in diabetes. This is one study I will remain skeptical of until more information is forthcoming.

I also don’t want it to sound like I am poo-pooing research. All dietary research is flawed, since we can’t do experiments in diets on people, it’s unethical. So to make up for the flaws, we have to keep doing lots of research, collecting lots of data, and spending lots of time analyzing and reviewing. Any information is better than no information.

I just had a great thought- what if they asked people to volunteer their information from all those diet/calorie tracking software programs? What a gold mine that would be! Plus, there wouldn’t be the inherent study bias, since the people would be turning over information they had previously gathered for themselves, with no unintended bias of pleasing the researchers.

Tag Cloud

%d bloggers like this: