You're absolutely right about the bad numbers. But wrong about "accuracy relies on representativity". A random sample of just one person is representative, but wildly inaccurate (large error, sampling variance). A non random sample of 1000 persons might be accurate, but only with regards to these 1000 persons, it is not representative (i.e. not representative of a larger population about which you would like to make deductions using the sample).
It is a common misconception that representativity stems from a large sample size, but it can only be the results of random sampling.
Actually, things are worse off then that... statistically speaking anyway. This survey (just like the/. or any other internet polls) is not representative because the sample is not a random one!
It is not even a random sample of geeks or/. readers. All you can do with these answers is draw conclusions about the people who gave them, you can't deduce anything about the larger population (may that be geeks, ipod owners,/. readers, your city, state, country or the whole world!).
The problem is not of accuracy, but of representativity.
Slashdot Mirror
You're absolutely right about the bad numbers. But wrong about "accuracy relies on representativity". A random sample of just one person is representative, but wildly inaccurate (large error, sampling variance). A non random sample of 1000 persons might be accurate, but only with regards to these 1000 persons, it is not representative (i.e. not representative of a larger population about which you would like to make deductions using the sample).
It is a common misconception that representativity stems from a large sample size, but it can only be the results of random sampling.
Actually, things are worse off then that... statistically speaking anyway. This survey (just like the /. or any other internet polls) is not representative because the sample is not a random one!
/. readers. All you can do with these answers is draw conclusions about the people who gave them, you can't deduce anything about the larger population (may that be geeks, ipod owners, /. readers, your city, state, country or the whole world!).
It is not even a random sample of geeks or
The problem is not of accuracy, but of representativity.