doug: (Default)
From: [personal profile] doug
This is less bad than the statistic makes it look, I think. The "1 in 4" figure invites you to imagine that this is the proportion of outright fraudulent statistics in published papers, but it's nowhere near that.

This data is a sample of consulting statisticians. They will have talked to a *lot* of scientists, and the question asked was a "have you ever" question. On the other side of the equation, people requesting statistical dodginess are likely to get turned down, and will then go on ask to another statistician, and another, and another, until they mend their ways, give up in exhaustion, or find one who gives in. That will mean a small handful of dodgy scientists are likely to contact a very large number of consulting statisticians.

(I'm not a statistician, but I have taught stats, and am probably number three choice for answering stats questions in my dept, and I get *loads* of people asking me for advice. I'm also good friends with the number one and number two choices and sometimes share stories when someone does the rounds because they didn't like the answer one of the others gave because it meant more work.)

The positive way of framing this study is "three quarters of people who spend their time answering questions from scientists who don't have enough grasp of statistics to do it themselves have never even been asked to do things that would, if done, count as fraudulent." I'd be surprised if a similar study didn't reach similar conclusions among, say, professional auditors - and they are on the hook for fraud in ways that consulting statisticians wouldn't be.

I'm not saying it's not disappointing and depressing, and I'm certainly not saying that there are not serious, widespread problems with the use of quantitative data in academic publications, but it's not nearly as bad as the "1 in 4 figure" instinctively makes you think it is.

February 2019

3 4 5 6 7 8 9
10 11 12 13 14 15 16
17 18 1920212223

Most Popular Tags

Style Credit

Expand Cut Tags

No cut tags
Page generated Feb. 20th, 2019 03:53 am
Powered by Dreamwidth Studios