Abusing statistics for fun

[Thudlow_Boink]

@Gedd

I prefer my kid watching Elmo over Spongebob.

Sorry, Elmo likes to be on the bottom.

[Gedd]

@Thudlow_Boink

Sorry, Elmo likes to be on the bottom.

:eek: Hey, don’t dis the red guy! Just yesterday I saw him do a mean rendition of “Hammer Time.”

[phungi]

When teaching stats, I always have fun writing:

19
__

95

on the board, and saying “you often get the right answer with the wrong solution”

and then cross out the 9’s

[RedSwinglineOne]

@Markxxx

As Archie Bunker said: “Ninty Five percent of all needless operations are totally unnecessary”

Gloria: Do you know that sixty percent of all deaths in America are caused by guns?

Archie Bunker: Would it make you feel any better, little girl, if they was pushed out of windows?

[Fotheringay-Phipps]

Here’s an example of the issue noted above:

Ioannidis’s first targets were shoddy statistics used in early genome studies. Scientists would test one or a few genes at a time for links to virtually every disease they could think of. That just about ensured they would get “hits” by chance alone. When he began marching through the genetics literature, it was like Sherman laying waste to Georgia: most of these candidate genes could not be verified. The claim that variants of the vitamin D–receptor gene explain three quarters of the risk of osteoporosis? Wrong, he and colleagues proved in 2006: the variants have no effect on osteoporosis. That scores of genes identified by the National Human Genome Research Institute can be used to predict cardiovascular disease? No (2009). That six gene variants raise the risk of Parkinson’s disease? No (2010). Yet claims that gene X raises the risk of disease Y contaminate the scientific literature, affecting personal health decisions and sustaining the personal genome-testing industry.

Statistical flukes also plague epidemiology, in which researchers look for links between health and the environment, including how people behave and what they eat. A study might ask whether coffee raises the risk of joint pain, or headaches, or gallbladder disease, or hundreds of other ills. “When you do thousands of tests, statistics says you’ll have some false winners,” says Ioannidis. Drug companies make a mint on such dicey statistics. By testing an approved drug for other uses, they get hits by chance, “and doctors use that as the basis to prescribe the drug for this new use. I think that’s wrong.” Even when a claim is disproved, it hangs around like a deadbeat renter you can’t evict. Years after the claim that vitamin E prevents heart disease had been overturned, half the scientific papers mentioning it cast it as true, Ioannidis found in 2007.