Replies

I use statistics all the time, professionally, for formal experiments on medical devices. The resulting documents sometimes end up being submitted to the FDA. So I know my stuff.

Do not pass go, do not hem and haw...whatever conclusions you draw are invalid. Any statistician would laugh you out of their office.

When you do a “random” sampling (polling of likely voters) of a population (all potential likely voters), then later find out your sample was actually biased (lots more dems than pubs), you must throw the results out...but the pollsters try to rationalize the imbalance by making it part of the "science" of what they do - saw a guy who helps do their polls on Fox last night - says he gets questions about the imbalance all the time - claims it's part of what might be called the dynamic party identification process - claims that party identification actually fluctuates with the fortunes of various candidates, so if pollsters get many more dems in a random sample than would be expected by normal patterns it's justified to use the sample because it just shows dems are doing particularly well at that time with the voters - in addition to the folly of the circular reasoning involved (we got more dems so dems must be doing better so we can include more dems because dems are doing better), his argument comes off the rails because he admits that most people's party identification doesn't change that much (lots of people say they're voting a certain way because their parents voted that way) but says there are "weak-identifiers" whose identification is fluid and that justifies changing the mix of dems/reps without referring to a fixed universe like voting patterns. But it's unreasonable to think this happens much at all in those they select to include in their polls since they insist they're using "likely voters" who could be expected to be much more consistent and committed to one point of view and party ID in their voting patterns. In any event - these pollsters should be kicked out of the office....