Flutterby™! : made up numbers 2007-07-05 23:30:50.25747+00

made up numbers

2007-07-05 23:30:50.25747+00 by Dan Lyke 8 comments

I've been fascinated by made-up statistics ever since Marc Klaas lied to congress about the alleged "50,000 annual non-custodial child abductions" (the real number is a few hundred). So what do we find when scientists debunk the "women talk more than men" myth? That, once again, the bullshit numbers are made up by the governmental control freak types:

That's because no original scientific sources exist for such a claim, said Mark Liberman, a professor of linguistics at the University of Pennsylvania.

His own inquiry found the earliest use of the figures by James Dobson, founder of Focus on the Family, who employed it as a "guesstimate."

"My current belief is that this started as someone's idea of a plausible estimate, and was turned into a false scientific factoid by writers who like to misuse the authority of science," he said.

Relatedly, on a drive down to a party yesterday, Charlene was reading a book on child psychology and we both got to taking apart some of the both obviously bogus claims, and the claims that couldn't be internally supported by the argument of the book. Damn, there's a whole lot of intentional fuzzy thinking out there.

This is the hallmark of scientific inquiry - "transparency." Unfortunately, the current model also relies on "authority". Frequently, the opinions of "authorities" get mistaken as measured facts and people are too lazy to do the literature review necessary to verify the facts. Moving the scientific body of literature to the web so it can be easily indexed and searched will be a big help.

Just listen to some right-wing talk-radio with a logical fallacy scorecard. You can usually bingo within a half-hour.

I would tend to think that would be true for talk radio in general regardless of the political bent.

Yeah, I tried listening to "Air America" for a while but found that, for the most part, it hd about the same attention to detail and fact as Rush. Entertaining, and sometimes in a "booyah" sort of way, but fell apart when faced with critical thinking.

As soon as you realize Rush is a very well paid entertainer, and that's ALL he is, he starts to make lots of sense.

#Comment Re: Made up numbers made: 2007-07-06 13:16:51.235588+00 by: m

Having had young children at the timeThere were a lot of numbers floating around back then. Another set said that 50,000 children a year were murdered. I guess 50,000 was a popular number at the time. There were fewer than 50,000 murders total, of which 500 were children. Total abductions figures of 2-3 million per year were also flogged. Sometimes such numbers were directly distributed by local law enforcement. The FBI and other experts would literally stand behind purveyors on podiums, lending their imprimatur without actually making the statements.

Other examples of fictitious statistics include the number of smoking deaths that Surgeon General Coop pulled out of the air at one Q&A session. A similar number for deaths from obesity, later bumped up by a third for no known reason. The introduction of BMI, a particularly inane method for categorizing obesity. The publishing of wild numbers for the incidence of HIV infection, that were well known to be false by the Public Health community.

Figures more likely to be familiar to the readers of this blog include the ridiculous amounts claimed to be lost to copyright infringers. Large software houses, the BSA, the MPAA and RIAA often claim (to Congress and elsewhere) that their losses are phenomenally large. The numbers are accepted at the Dog and Pony shows, the malfeasant "laws" are passed, and the Congress persons go home with pix and autographs of the celebrities trotted out, as well as the "contributions."

How many hundreds of thousands of Iraqis was Hussein supposed to have killed, even the the mass graves were as scarce as WMD?

Another big problem with academics is that they tend to horde data. They will get a grant to do some research and then will keep the original data secret for as long as possible to generate as many publications out of it as possible. Some organizations, like the NSF, have a data publishing requirement. But that data isn't necessarily published in an accessible manner.

The hording of data, including and especially that which is collected on the public's dime, is a serious issue that doesn't seem to have been resolved. This is even more true when the claimed analytic results seem to demand a significant change in public policy.

CDC is good, though not infallible. There are potential issues with CIA data, but it is probably the best available for that genre. One of the things it took a long time to truly learn, is that no one's data (especially my own) is anywhere near as good as the owner thinks it is. And, it is important to remember that recent metastudies show that more than half of all medical publications are wrong -- what ever that means.

The first step is really to decide what numbers need to be questioned. To develop a rapid screening test for reasonability. Many people can not do this, but I am not sure why. Are they intellectually incapable, or has such a thought process never been taught to them?