Benford's Law: faking "random" numbers

[cuterebra]

Humans aren't very good at generating truly random strings of numbers. Neat article about how to identify fraud.

http://www.rexswain.com/benford.html

 

[Da Blob]

Humans aren't very good at generating truly random strings of numbers. Neat article about how to identify fraud.

http://www.rexswain.com/benford.html

Good article. It so happens that fraudulent researchers are not very adept at faking 'random' samples either, it would be nice to develop software to detect such - but then there would be a lot of peer-reviewed journals that would have to cease publishing...

 

[Aiss]

^ It has been in use for some time now... From wikipedia:

Evidence based on Benford's law is legally admissible in criminal cases at the federal, state, and local levels.http://en.wikipedia.org/wiki/Benford's_law#cite_note-6

My teacher did similar thing with flipping the coin homework in probability theory class.

 

[Misanthropy]

I just saw this article on reddit today, really interesting.

 

[Nicholas A. A. E.]

This is really interesting. It seems there should be a simple explanation for the prevalence of Benford's law, but apparently there isn't. A complex one, yes, but that's not what i would expect.

 

[cuterebra]

This is really interesting. It seems there should be a simple explanation for the prevalence of Benford's law, but apparently there isn't. A complex one, yes, but that's not what i would expect.

I had the same thought, initially--Occam's razor and all.

 

[Nicholas A. A. E.]

Well, I think Occam's razor is absolute nonsense the way most folks today tend to use it. But that's a different discussion.

 

[Architectonic]

Certain numbers, such as the number one hold psychological importance, such that most of our units are scaled so that the number one is more frequently dealt with than other numbers.

 

[Murphy1d]

Well, I think Occam's razor is absolute nonsense the way most folks today tend to use it. But that's a different discussion.

I hear you, when statistics and complex principles are being used by people who don't truly understand them I get a headache.

Don't get me started on "random sampling." All knowledgeable textbooks state (in essence) "use sampling guidelines and practices, but only while it makes sense."

Thus, if you want to improve process quality and your process is 92% electronic (and 99% accurate) and 8% manual (and 75% accurate) don't make your sample size 92% electronic items and 8% manual items.

Focus on the variation (the 25% errors).

 

[Nicholas A. A. E.]

One hilarious mistake in my psychology textbook implies that random sampling automatically gives you representative samples. Absolute nonsense, of course: random samples have no particular tendency to be representative or non-representative. Depending on how you arbitrarily designate "representative."

Little things like that every now and then really lower my respect for psychology as a science.

 

[typondis]

The article was not about detecting fraud. It talked about the use of Benford's law in doing that, but mostly it was about number sequence, distribution, and occurence.

Besides, in a Leisure State there would be nothing to be fraudulent about.