----- Original Message ----- 
From: "JDG" <[EMAIL PROTECTED]>
To: "Killer Bs Discussion" <[EMAIL PROTECTED]>
Sent: Tuesday, October 12, 2004 6:24 AM
Subject: RE: 2004 Presidential Race Analysis


> At 07:10 PM 10/12/2004 +1000 Andrew Paul wrote:
> >> From: Erik Reuter [mailto:[EMAIL PROTECTED]
> >> The data is really too close to call. People can analyze all
> >> they want, it doesn't change the fact that the polls are
> >> within the margin of error.
>
> Using a large number of polls can substantially reduce the number of
error.
>
> That is, while one poll of 50-46 would be "within the margin of error" -
> having six such polls would lead you to say with some confidence that
> candidate A is slightly ahead, probably by around 50-46.
>
That is true if and only if the errors can be treated as statistical or
pseudo-statistical.  If there is a small systematic error that is shared by
the polls (by small I mean 1%-2%).

Most polls just before the 2000 election showed such a bias.  IIRC, Zogby
was the one poll that was close to on top of the right number. IIRC, that
was true in the previous election, tool.

Dan M.


_______________________________________________
http://www.mccmedia.com/mailman/listinfo/brin-l

Reply via email to