Brandon and Chris both made good points. Another way to look at this is to note that the question, "What if you are one of the two?" has an implied sense that if you were tested you'd be identified and therefore saved- 7 - 2 = 5. But that assumes that "the two" who are going to die of the disease would be "caught" by the testing. Clearly not so. "What if you were one of the 5?" is a more telling question. Clearly being tested isn't going to help them. And there is no way to know prior to testing that the 5 are the 7 minus the two, as it were. The statistics, as Chris' analysis has shown for several of these examples of causal confusion, don't work that way. Tim _______________________________ Timothy O. Shearon, PhD Professor and Chair Department of Psychology The College of Idaho Caldwell, ID 83605 email: tshea...@collegeofidaho.edu
teaching: intro to neuropsychology; psychopharmacology; general; history and systems "You can't teach an old dogma new tricks." Dorothy Parker -----Original Message----- From: Christopher D. Green [mailto:chri...@yorku.ca] Sent: Tue 4/14/2009 7:41 PM To: Teaching in the Psychological Sciences (TIPS) Subject: [tips] [Bowel cancer risk in older women could be reduced by more soya in diet | Science | The Guardian] Michael Smith wrote: > > Hmmm. > But what if you are one of those two? Then you die (by definition). The problem is you don't know you are one of those 2 until you are dead. We have to make these decision on the basis of the probabilities well ahead of time. There are lots of things that raise our chance of death by a lot more than 2 in 10,000, and we should worry about them first. Smoking in #1. Smoking raises the chances of a 65 year old man dying of lung cancer in the following 10 years from 4 in 1000 to 89 in 1000. Now there's some action. This 2 in 10,000 stuff is mere peanuts in comparison. I don't have then numbers, but I'll bet you $5 that talking on a cell phone while driving increases the probability of having a car accident by a lot more than 2 in 10,000. (According to some research, talking on a cell phone makes the driver as impaired as drinking over the .08 level.) Chris Green York U. Toronto ============= > > > On Sun, Apr 12, 2009 at 10:31 PM, Christopher D. Green > <chri...@yorku.ca <mailto:chri...@yorku.ca>> wrote: > > > Here's another example of misleading medical statistics example > you might want to use in class. > http://www.guardian.co.uk/science/2009/apr/11/bowel-cancer-soya > > The report says that more soy in the diet (it doesn't say how much > more) can reduce women's chance of death from bowel cancer by 30%. > (Actually, it says those in the top third of soya intake had a 30% > decrease in bowel cancer deaths, compared to the bottom third -- > NOTE: not from the national average -- but it doesn't give any > indication of how much soy each group actually ate.) > > So let's work the numbers. 16,600 women die from bowel cancer per > year in the UK. There are about 60 million people in the UK. Half > are female: 30 million. Lose the 20% children and we have 24 > million women. > > So, the chance of dying of bowel cancers for women in any one year > is: > 16,600/30 million = .0007, or 7 in 10,000. > > A 30% reduction would lower that chance to 5 in 10,000. > > So (even if the causal implication that is not actually > demonstrated here were correct), if increasing your soy intake by > a fair bit could decrease your chances of dying by bowel cancer by > 30%, that would represent a tiny reduction of just 2 in 10,000 per > year. > > > Chris > -- > > Christopher D. Green > Department of Psychology > York University > Toronto, ON M3J 1P3 > Canada > > > > 416-736-2100 ex. 66164 > chri...@yorku.ca <mailto:chri...@yorku.ca> > http://www.yorku.ca/christo/ > > ========================== > --- To make changes to your subscription contact: Bill Southerly (bsouthe...@frostburg.edu) --- To make changes to your subscription contact: Bill Southerly (bsouthe...@frostburg.edu)
<<winmail.dat>>
