On Oct 12, 2006, at 2:47 PM, ravi wrote:
> Though if at the end of the day you cannot compare these numbers, then
> that questions the very statistical validity of their results.
No it doesn't. Questions are asked in different ways, and different techniques are used to inflate the sample into a representation of a full population. The NYT/CBS poll, for example, consistently gives Bush a lower approval rating than Gallup. Both move in tandem, but typically with a gap of 3-4 points. But if you used a Gallup poll for Sep, which showed Bush with a bounce to 44%, and a NYT/CBS poll for Oct, which shows 34%, you'd exaggerate the decline in Bush's rating.
Doug