3 ms·
See also Gerd Gigerenzer's book "Reckoning with Risk". https://plus.maths.org/content/reckoning-risk https://plus.maths.org/content/reckoning-risk >The probab
by DanBC 11y ago
See also Gerd Gigerenzer's book "Reckoning with Risk".
https://plus.maths.org/content/reckoning-risk https://plus.maths.org/content/reckoning-risk
>The probability that one of these women [asymptomatic, aged 40 to 50, from a particular region, participating in mammography screening] has breast cancer is 0.8 percent. If a woman has breast cancer, the probability is 90 percent that she will have a positive mammogram. If a woman does not have breast cancer, the probability is 7 percent that she will still have a positive mammogram. Imagine a woman who has a positive mammogram. What is the probability that she actually has breast cancer?
> Eight out of every 1,000 women have breast cancer. Of these 8 women with breast cancer, 7 will have a positive mammogram. Of the remaining 992 women who don't have breast cancer, some 70 will still have a positive mammogram. Imagine a sample of women who have positive mammograms in screening. How many of these women actually have breast cancer?
His point is that it's not just patients who don't know the numbers. Doctors and nurse, who use these numbers every day, also don't understand them.
- Houshalter 11y agoThat's a classic problem to introduce Bayes theorem. However the more interesting fact is 95% of doctors get that question wrong.
- incompatible 11y agoSurely doctors would learn to expect the numerous false positives after a while. Or does the question not really relate to reality?
- pjscott 11y agoThe "imagine 1000 women..." formulation definitely sounds easier to understand, but doctors do very poorly on it as well. Some numbers: http://slatestarcodex.com/2013/12/17/statistical-literacy-among-doctors-now-lower-than-chance/ http://slatestarcodex.com/2013/12/17/statistical-literacy-am... In practice things are not quite as dire as you'd expect, thanks to diagnostic guidelines written with input from people who know about conditional probability.