|
Edited on Mon Apr-25-05 09:23 AM by TruthIsAll
There are three types of weightings made to raw data. If there are others, please let us know.
The first makes sense. The second does, but only if... The third makes no sense.
1- Weighting the exit poll sample to proportionately match the NUMBER of voters.
This makes sense.
For example, there were 872 CT respondents (of the 16261 total regional respondents, or 5.363%).
There were 1.551 million CT votes out of the 26.854 million who voted in the East, or 5.776%.
The weighting factor WF = 5.776/5.363 = 1.077
Kerry won CT Exit Poll with 58.47% (510 of the 872 Exit Poll respondents).
Kerry's re-weighted exit poll percentage is: KW = KP * N * WF KW = 58.474% * 872 * 1.077 = 549
2- Weighting to match the voter DEMOGRAPHICS (Female/Male = 54/46$). This makes sense, assuming the weightings TRULY reflect reality.
For example, we know that it was IMPOSSIBLE for 43% of 2004 voters to have voted for Bush in 2000. The MAXIMUM weight possible was 39%. So in this case the weighting makes no sense.
3- Weighting to match the recorded vote percentages. Kerry's share of NEW voters was adjusted from 57-41 of first 13047 respondents to 54-45 for the 13660 Final. How could the 613 final exit respondents cause this? In fact how could these mysterious 613 cause a Kerry 51-48 win (based on the first 13047) to a Bush 51-48 win?
This makes NO sense to me.
|