Wednesday, October 06, 2004
CBS News | Uncommitteds Tab Edwards Winner | October 6, 2004?07:52:39
The sample size is 178 "uncommitted" voters. How many people needed to be contacted to arrive at 178?
A back of the envelope calculation would start with the number of uncommitted voters? Is 20% too high? If not, then a quick figure would be 890.
My guess is Knowledge Networks used an operational defitnion that would easily define a voter as "uncommitted".
The organization could not have used a much smaller ficure as the number of respondants would quickly move into the realm of impausibility. 4% undecided would mean 4450 respondants were polled.
None of the major polls contain this many respondants. Most hover in the 1000 range over three days of polling (350 per day).
The last debate occurred less than a week ago. 5 days to be exact.
This smells like GIGO.
The sample size is 178 "uncommitted" voters. How many people needed to be contacted to arrive at 178?
A back of the envelope calculation would start with the number of uncommitted voters? Is 20% too high? If not, then a quick figure would be 890.
My guess is Knowledge Networks used an operational defitnion that would easily define a voter as "uncommitted".
The organization could not have used a much smaller ficure as the number of respondants would quickly move into the realm of impausibility. 4% undecided would mean 4450 respondants were polled.
None of the major polls contain this many respondants. Most hover in the 1000 range over three days of polling (350 per day).
The last debate occurred less than a week ago. 5 days to be exact.
This smells like GIGO.
Comments:
Post a Comment
