|
The E-M polls were not designed to detect the variability of interest at the unit of interest level. Why not? I joined several groups in analysis of data, and there seems to me there there is an issue.
When we have a "minor" problem in education or health...we target the assessments and sampling to detect the causes. Pollsters continue to take money (as I understand it) from news services and seem to relish in the controversy, but really don't have the goal of finding out the "true vote" as a goal?!?!
If E-M won't do the job, then let's get someone who will. We "assume" that the news services want to report the truth. Your (Feeble) analyses, as technically correct as they are, seem narrow in focus (hypothesis testing) and less descriptive or relational. Both ideas hold a solid place in the discovery of reality.
I can think of many designs that might discover fraud and manipulation. I'm just wondering why 2 or 3 or 4 elections occur that the pollsters don't try them? I don't care if it's complicated (multivariate) or sampling (targeted precincts) or many other solutions.
TIA overreaches the probabilities, and many articles, such as these threads, raise more questions than answers. There is NO QUESTION that 10 times the undervote in an isolated precinct like in Sarasota in the last election MUST be either serious machine error or fraud. Also, caging is illegal, and we have lots of evidence of caging in Fl that is CONSISTENT BUT NOT PROVED by exit poll data and eyewitnesses.
As a "researcher", I'm not immune to social consequences. I just finished a contract with NSF. If I discovered something that harmed people or the environment, I'd report it. If I suspect a legal or ethical issue, I'd design something to refute or confirm it as part of the research. Why doesn't E-M do that?
Sometimes OTOH seems to shoot the messenger. The problem of voting integrity can be solved. Would the entry point of exit pollsters be a logical place to start? If so, we need to hold them to the fire. They should a better job and report evidence that is consistent with the "real vote" or give the job to someone else. If there are inconsistencies, E-M should identify and explain them. The "reluctant responder" and "cute interviewer" stuff has gotten out of hand. If E-M can't resolve the questions, then I'm sure that Freeman and others would be glad to have the contract....
|