Reply #11: I'm glad that TIA raises the questions. A top 10 list.

[Sancho]

Even though poll analysis may not be statistically convincing in traditional terms (p<.05 and all that R. A. Fisher stuff), there are several positives to TIA's posts:

1.) Anecdotal evidence of vote switching, undervotes, etc. are pointing to election fraud, but not "legal proof", and that is consistent with TIA's hypotheses; a form of convergent validity
2.) Some of TIA's analyses reveal patterns that hold pollsters feet to the fire in the future and they may be forced to be more transparent about the popular processes
3.) If we think that elections can be rigged, a well-designed exit poll at the precinct level (parallel election) is likely the only way to provide clear evidence of all the possible manipulations of machines, tabulators, provisional ballots, etc. so that proper investigations and re-elections can be held
4.) Some of TIA's analyses are refutable, but as stated before, non-ignorable nonresponse can only be determined with heroic efforts (see Howard Wainer: Eelworms, Bullet Holes, and Geraldine Ferraro: Some Problems in Statistically Adjusting for Survey Nonresponse) and TIA is making the heroic effort that needs to be done - maybe something will surface that can't be refuted at some point
5.) TIA's posts have generated a number of other investigations by statisticians and political scientists - good for the order to do this thinking
6.) Even when a given analysis is questionable in sampling/probability terms, the variety of analyses as a group by TIA have more convincing weight - sort of a poor man's meta-analysis
7.) TIA's discoveries remind me of John Snow ( http://www.ph.ucla.edu/epi/snow/fatherofepidemiology.ht... ). Sometimes there is no scientific proof, but the pattern can show one where to look
8.) TIA has certainly got more people reading about polls and statistics that would normally be interested. A more sophisticated public won't be as easily fooled by unexplained weighting in the future
9.) TIA has been responding to criticisms over the last couple of years and takes the debate to a new level about WPE, reluctant responders, and attractive pollsters
10.) TIA's viewpoint is not one of an insider (pollster) and many disciplines can't see the forest for the trees, so having an outside examination is a good idea; the rhetoric that "exit polls can't prove fraud" is mostly a pollster's lament, not a universal law of physics!