In the fall of 1996, I was in a Health Physics class at Texas A&M University with a bunch of new faces. Early in the semester (it might have been the first day), we were discussing risk analysis and the actuarial risk of various activities. As part this discussion, a table from Bernard Cohen's book (Before It's Too Late: A Scientist's Case For Nuclear Energy) was showed to the class. The table was titled "Average Loss of Life Expectancy (LLE)" and here is an extraction of the highlights that Avery (one of my new classmates) thought applied to him:
Activity Causing LLE (Days of LLE)
- Being male (2800)
- Being unmarried (2000)
- Smoking cigarettes, 1 pk/day (1600)
- Being poor (700)
- 15 lb overweight (450)
- Accidents (435)
- Driving a car (20)
- Alcohol (130)
- Occupational Accidents (74)
- Small cars vs. standard size cars (50)
- Falls (39) [Avery was kind of clumsy.]
- Radiation worker entire adult life (12)
When Avery totaled up the days, he proclaimed that it was 21.5 years. At this point, he stopped the class by saying, "I can't handle this! Those numbers can't be true! If they are, I am living on borrowed time." These numbers have probably changed a little bit since Cohen's publication in 1983, but Avery's logic path was one that is commonly tread in medical science journalism today. However, I am pretty sure that he was just being funny.
We are bombarded daily with media reports of ecological type of epidemiological studies that tells us something like, "Men who eat green cheese are 21% more likely to need vision correction... Dr. Joe Blow is encouraged that green cheese lovers may save millions of dollars in eye care related expenses each year by simply moderating their consumption of this offensive dietary item. Dr. Blow also stated that more research is needed on this health care crisis." Like this story, almost every one of this studies reported on a daily basis is pure, unadulterated bunk.
Epidemiologists tend to believe that relative risks between 0.5 - 2.0 imply, at best, weak associations, and that studies such as the green cheese study above should be viewed with a great deal of suspicion because of the nature of this epidemiological method and the data quality that comes from ecological studies. (Ecological studies are observational studies where data is collected on populations rather than individual subjects.) This is true because statistical correlation doesn't mean that you have identified a true cause and effect relationship.
Steve Milloy has an excellent tutorial on how to recognize and debunk junk science, so I won't spend time trying to redo his work. I don't have definitive proof, but I think Mark Twain was referring to these "researchers" when he wrote that "There are three kinds of lies: Lies, damned lies, and statistics." What I find more than a little disturbing is that we are allow these grabs at taxpayer money by pseudo-scientists that abuse both true research methods and statistics. Until we can find a way to combat junk science, I guess I will have to find comfort in the fact that when I find purple wild flowers growing in the northwest portion of my freshly mowed yard, I really don't have a 4% increased risk of developing testicular cancer.