> Too many voting ’irregularities’ to be coincidence
11 November 2004, 02:05
The graphs contrast exit poll results to actual votes and suggest that only fraud could explain the discrepancies, but the evidence presented is highly selective.
Statisticians please comment (and state your professional qualifications): what kind of statistical comparison would be meaningful? If all the exit polls in the country were compared to actual results, couldn’t we determine the probability that a particular result in a particular precinct could have occured by normal variability, as opposed to error or fraud?
Pollsters please comment (and state your professional qualifications): some imply that the exit polls should be discounted because they are partisan or unscientific. Who carried out the exit polls - what polling companies in what states? It is my impression that polls are conducted by the major polling companies in a disciplined scientific manner and used to set the margin or errors, and that actual results outside the margin of error are suspect. True?
I believe the argument can only be won through statistical analysis.
The graphs contrast exit poll results to actual votes and suggest that only fraud could explain the discrepancies, but the evidence presented is highly selective.
Statisticians please comment (and state your professional qualifications): what kind of statistical comparison would be meaningful? If all the exit polls in the country were compared to actual results, couldn’t we determine the probability that a particular result in a particular precinct could have occured by normal variability, as opposed to error or fraud?
Pollsters please comment (and state your professional qualifications): some imply that the exit polls should be discounted because they are partisan or unscientific. Who carried out the exit polls - what polling companies in what states? It is my impression that polls are conducted by the major polling companies in a disciplined scientific manner and used to set the margin or errors, and that actual results outside the margin of error are suspect. True?
I believe the argument can only be won through statistical analysis.