|
|
Post by Tony Crispino on Jan 23, 2015 14:38:12 GMT -8
Well having completed my 2nd Cochrane Collaboration patient advocacy truing course. A new category I thought would be fine to add is Fun with Statistics. We tend to try to make heads or tails of complex medical statistics and many times get fooled by bias and marketing. A big thanks to Musa Mayer on her courses on an Evidence Based Healthcare System. Absolutely fun and interesting courses. Here are some basic rules on how to "talk back to a statistic" When you see a number, ask questions ... - Who says so?
- How does he/she know?
- What’s missing?
- Did somebody change the subject?
- Does it make sense?
My next posts will be a few examples... |
|
|
|
Post by Tony Crispino on Jan 23, 2015 14:46:12 GMT -8
The Framing Effect on Statistics.
Patients and medical professionals were asked to make a hypothetical decision between surgery and radiation treatment for lung cancer.
> Group A told that for every 100 men undergoing lung surgery, 90 will live through the post-operative period, and 34 will be alive at the end of 5 years
> Group B told that for every 100 men undergoing lung surgery, 10 will die during the post-operative period, and 66 will be dead by the end of five years
More in Group A chose the surgery
. What did you think?
The statistics are identical. Both a and B are exactly the same statistics inverted. You may have seen it before. Comparing radiation or radical prostatectomy for prostate cancer therapy if you go to one bias the provider may describe HIS treatment like it is in group A above and might describe other options like group B.
And vice versa...Be Aware of statistical framing...
|
|
|
|
Post by Tony Crispino on Jan 23, 2015 14:56:57 GMT -8
Non-Cancer but marketing related:
Shortly after the September 11, 2001 terror attacks, the NTSB issued a diagram showing the following statistics:
Automobile Deaths 1999-2000.............41,780
Commercial Airline Deaths................... 92
Boeing Corporation latched on to it and had it on their website for many years stating how much safer you are traveling by commercial airlines than by car. But Boeing had a huge bias and much is left out of the data. For Example:
There is no common denominator applied to the above statistics. What might be needed to compare:
> There were 13,000 commercial airlines aircraft in the US in 1999-2000
> There were 140,000,000 cars on the road in 1999-2000
To determine risk your would need to compare how much time people spent flying versus riding in a car. How many times an aircraft takes off and lands. After crunching the numbers the risk is in flying with a 5 times more likelihood of dying if equal time was spent doing both.
|
|
Walt Shiel
New Member
Completed HDR + IMRT/IGRT: 10/17/14
Posts: 7 
|
Post by Walt Shiel on Jan 29, 2015 11:50:28 GMT -8
Tony,
Great idea for a topic. As an engineer with considerable background in number-crunching large amounts of data, I know my BS flag flies high whenever somebody starts throwing statistics at me. I want to see the raw data, not just those capsules of, all too frequently, carefully massaged data we call statistics.
If I analyze a 10-year slice of data (it doesn't matter on what topic) while ignoring the 100 years on either side of that slice, I may easily derive a totally invalid statistical result. Likewise when evaluating a population of people for any particular trait of reaction.
We need to know the assumptions and selection criteria for those statistics before giving them any real credence.
Walt
|
|
