[Edu-sig] Talking Stat Blues

[Arthur]

Jason writes -

>I can't take it any more. I really need your help please..

>In these painfully contorted times, media are forever saturating news
>reports with the latest polls just in.

>53% think yes leader x is lying scum
>46% said no he's not, he's got Dog on his side
>1% undecided
>etc 

>If you are very lucky they specify the size of the sample. 1000 people
>perhaps.
>But they never speak it. Instead we get all this nonsense 

>"Most Martians said ....."
>"The majority ...."
>etc

>I need your help to find a [new] more honest, preferably Pythonic and
>revealing way to speak about these numbers.

Jason, you sure expect a lot of Python.

But as it happens, I understand that Python 2.4 will include a statistics
module. But I think you will find it to be part of the problem, rather than
part of the solution.  As I am assuming it implements and makes available
the functionality that allows one to draw conclusions, for a large
population, within a defined margin of error, from a small sample.

It seems to me the study of such a module would lead you to conclude, with
me, that the distortions you sense may exist in such polls are in the words,
not the numbers.

Art