At 17:10 09/11/00 -0600, mwang wrote:

>When the presidental poll results were shown on TV, a margin error of 3 or 4 >was often reported. How was this computed, based on the sample size, or based >on percentage result?

Theoretically both - but, in practice (for elections with two predominant candidates) essentially sample size. The 'margin of error' for a given sample size varies little over the range of 'percentage results' in the range, say, of 20% to 80% - hence, in practice, it's really only the sample size that is relevant for these type of polls.

A (95% confidence interval) margin of error (for the percentage vote of one candidate) of ('plus or minus') 3% or 4% for for results in the 40-60% range is quite commonly seen in political polls. Those margins of error corresponds to a sample sizes of about 1000 and 600 respectively. 'Plus or minus 2%' is also commonly seen, and that corresponds to a sample size of around 2400.

The margin of error for the score of an 'outsider' candidate expected only to get a few percent of the votes would be appreciably smaller than for the main contenders.

Kind Regards,

John

---------------------------------------------------------------- Dr John Whittington, Voice: +44 (0) 1296 730225 Mediscience Services Fax: +44 (0) 1296 738893 Twyford Manor, Twyford, E-mail: John.W@mediscience.co.uk Buckingham MK18 4EL, UK mediscience@compuserve.com ----------------------------------------------------------------