Re: Inaccurate Statistics (Agreed chivo243)
Using the USA as an example, 535 divided by 310 MILLION is 0.000001726.
It's a survey of workers, so your denominator should be more like 147 million. If you're going to bitch about their methodology, you might at least try to get the numbers vaguely right.
And, of course, even that number is wrong, because what the survey is actually studying is perceived productivity and work habits of salaried white-collar workers. Non-ag wage and salary employment is around 136M. Part-timers account for around 19% of that, so now we're down to ~110M. I haven't found a handy breakdown of those remaining 100M workers, but clearly you were off by more than a factor of 3.
More importantly, if certain conditions hold, the required sample size to achieve a given margin of error and confidence level doesn't grow with the population size. Once the population grows beyond around 20000, a sample size of about 400 remains sufficient for a 5% margin of error and a 95% confidence level (p<0.05).
Now, those conditions may be difficult to achieve in many situations, and many statisticians are dubious about this sort of polling in particular. The necessary conditions include a homogeneous population for the conditions you're testing for, and a selection process that chooses an unbiased-random sample from the entire population you're claiming to examine. And then polls of this sort suffer from all sorts of other methodological problems: self-reporting, subjective questions, and so on.
But - and this may come as a shock to many Reg readers - there are actually some people who have put some thought into these problems. Your cunning and trenchant critiques are not novel! I know the Reg Brain Trust considers itself on the forefront of every field of intellectual inquiry, but on the rare occasion specialists have gotten there first.
Biting the hand that feeds IT
