Derb: If the GSS had some decent graphing tools built in instead of those clunky bar charts, you'd be seeing a lot more of this.

There is a bit of a learning curve, though granted not a terribly steep one - but between getting used to searching for variables, selecting on Rows vs. Columns vs. Controls and the filtering syntax, and then how exactly to go about capturing and editing the graphs/tables you want to display, it can be a bit daunting until you get your feet wet. You can probably count on dedicating a good hour or so the first time GSS-ing before getting anything useful out of it.

Poking through the GSS is actually very and doesn't take much time when compared to spinning theories based on personal intuition. So why is the latter still so much more popular than the former?

They are not used by people in the same industry. There is the figure-shit-out industry, which is data-driven. They don't pretend to have some crystal ball -- they take a look at some data and give their impression, hopefully a good one. If they're trying to show off any skill at all, it's their ability to analyze data. They'll take a look at all sorts of data.

Then there is the soothsayer industry, for which data are superfluous. The point is not to figure out what's going on -- it is to show what an awesomely insightful vision of the universe the speaker has. Of course, they're always full of shit, but the morons (everyone) never seem to notice because they prefer the taste of charismatic storytelling over plausibility and accuracy when digesting a plate of thoughts that an expert has prepared for them.

Unlike the data people who examine all sorts of stuff and build off of their learning, the soothsayers have a single grand vision that is already formed in their brain, and that permeates their insights into any topic, even a new one with no relevance to the theory -- Islamofascism, class struggle, or whatever.

Philip Tetlock has a big book, Expert Political Judgment, showing how the soothsayer pundits are worse than chance at predicting -- which means they don't even understand the present or past. The mostly an-ideological data people are not so bad.

So, most people don't use the GSS because they simply are not curious about how the world works. (Some don't because they are but don't want to learn the interface.)

Writing a book that synthesizes a bunch of GSS results (or stuff from the Statistical Abstract of the United States, or whatever) is pedestrian -- glorified clerical work, with some two-bit reporting. Writing a book called The Lexus and the Olive Tree or The Population Bomb is not risible, but proof of your penetrating insight into the fabric of the cosmos, and rich idiots will pay you thousands (maybe tens of thousands?) of dollars to convey your vision to their corporate audience.

Pretty easy career call if you're looking for a comfortable life.

Philip Tetlock has a big book, Expert Political Judgment,

lol. i was thinking of this book after your first paragraph. but right, they're not analysts, they're entertainers. it's kind of like how taco bell is not mexican food, it's mexican-themed food.

Whiskey/testing99/evil neocon said something about men vs women's views of immigration that sparked some spelunking on my part in the GSS here. The Audacious Epigone gave some helpful advice I'll try to take in the future. As a bonus for CS nerds, I get into Big-O growth in the comments.

Bryan Caplan defends experts in his review of Tetlock's book appearing in Critical Review, which I provide here.

By the way, in looking for good posts of yours for Quarks Daily one that stayed in my mind was the one on D.J Drummond (even if not appropriate for the competition). That's what happens when you say "screw the numbers, I KNOW that I'm right!".

Also, speaking of stats, does anybody have any ideas on the best way to investigate this?

Not to defend ignorance or refusal to test - but questionnaire surveys are regarded as a pretty crumby source of data in medical science - and the idea that a single survey could *resolve* a scientific question is a joke.

As a rule of thumb, it is best to use evidence from a variety of domains or methodologies in tackling a scientific question.

If the question is psychological the relevant evidence might well include many types of lab-type science - both animal and human - as well as drug studies, medical data, ethnography ... it's hard in advance to say exactly what might turn out to be relevant.

Think, for instance of the range of information brought to bear in analyzing IQ, by someone like Geoffrey Miller - who is now studying sperm!

http://www.economist.com/ science...ory_id=12719355

[P.S 'Ms Arden' is Miller's wife]

on D.J Drummond (even if not appropriate for the competition).

funny you bring him up, i recall bruce (whose comment follows yours) defended the moron drummond in outranged tones ;-) of course vindication was sweet since ben showed that yes, the polls were right, because you know there was actually a vote. so it seem's bruce's requirements for methodological rigor vary.