Email this thread to a friend
Bookmark this thread
|Home » Discuss » Topic Forums » Election Reform|
|Febble (1000+ posts) Send PM | Profile | Ignore||Sun Feb-11-07 12:08 PM
Response to Reply #77
Edited on Sun Feb-11-07 12:36 PM by Febble
This is bullshit. I have addressed every-one of these assertions of yours, countless times. I don't know whether you simply ignore my responses, or fail to understand them. I will address them again, only once. From this point onwards, I will simply link to my rebuttals:
You believe there was fraud, but do not believe it can be detected in the pre-election and post-election polls.
I don't know if there was fraud or not. I do not believe that the pre-election or the exit polls are evidence of it. I think that the exit poll data if anything suggests that it did not happen on a scale of millions.
You believe in scieentific polling, but do not believe the results.
I know a fair bit about survey research, and like all involved in survey research I am aware that non-sampling error can be substantial, particularly when face-to-face selection methods are used, and where non-completion rates are high.
You believe in statistical analysis, but not when it indicates that Kerry won the election.
I believe in sound statistical analysis, whatever it indicates. I don't consider yours sound.
You believe in the Law of Large numbers, but not when it applies to multiple pre-election polls taken for the same population at the same time.
As I said, the Law of Large numbers refers to random samples. If you still think that the only error variance in surveys is sampling error then you have learned nothing over the last two years.
You believe in probability analysis, but not if it indicates that Bush stole the election.
You believe in playing "what-if" sensitivity analysis (or as you prefer, hypothesis testing), but not if the analysis shows Kerry to be the winner of all plausible (and some implausible) scenarios in which vote share and turnout assumptions favor Bush.
My evaluation of any test of any hypothesis has nothing to do with whether or not I happen to like the result. It has to do with the validity of the assumptions. I do not share your assumptions. Your implication that my interpretation of data depends on whether I like the result is offensive.
You believe in regression analysis, except if it shows that Kerry did better in states with low exit poll response and Bush did better in states of high response.
I believe in interpreting regression analyses coherently. I don't know what you are inferring from this one, but it makes no sense. What you probably mean is that redshift tended to be uncorrelated with response rate. This is true, and some have tried to use this finding as evidence that non-response bias cannot have been the cause of the exit poll discrepancy. Unfortunately, this argument does not bear close scrutiny, and is falsified by the data. The reason it does not bear scrutiny is that it is an example of the ecological fallacy. Bias will only occur if there is a discrepancy between the response rates of Kerry voters and the response rates of Bush voters. It doesn't matter what the overall response rates are, and these may be determined by quite different factors to those that are associated with differentials between Bush and Kerry response rates.
You don't believe in faith-based analysis, but you disregard the fact that 43 states deviated from the exit polls to Bush.
This is a totally false statement, and is evidence that either you do not read my posts, or you willfully interpret them.
You don't accept that the pre-election polls matched the exit polls, even when I presented the data in this thread which proves it.
You still fail to understand the point made by both Lindeman and myself that the magnitude of the discrepancy between the offical results and the exit polls is completely uncorrelated with the discrepancy between the official results and the pre-election polls, even using your own very generous (to Kerry) pre-election poll estimates. If things are completely uncorrelated, they cannot, in any reasonable sense, be said to "match". If you want to re-invent the English language, feel free, but warn us that you are speaking something other than English.
You claim to be a scientist, but you cling to an unproven theory of false voter recall based on a single NES study to explain the exit poll discrepancies. That is very unscientific. You are excluding the best evidence - which is who the respondents actually said theyvoted for just FIVE minutes after voting.
TIA, think. Of course I am not "excluding" evidence as to whom voters said they just voted for. What I am regarding with some circumspection is the evidence as to who those voters voted for in 2000. Either you are deliberately misrepresenting my position or you have still failed to comprehend it.
You claim to have seen the raw data which indicate that Bush won, yet the Ohio ballot data which has been made available to investigators such as Richard Hayes Phillips, Fitrakis and others indicate just the opposite.
I have no strong views on whether Bush or Kerry "won" Ohio, which I consider a thoroughly corrupt election. I am so far unpersuaded that the evidence suggests that Kerry would actually have won on a level playing field, but I certainly consider it possible. The fact that we don't know is itself a scandal. What I do say is that the exit poll evidence in itself is not evidence of fraud in Ohio.
You know that Bush stole the election from Gore in Florida, yet you claim that he won the state in 2004, despite the University of Berkeley study which corroborates the documented fraud.
Oh, for goodness sake. You do know, don't you, that that analysis has actually been withdrawn? It was thoroughly rebutted. There is as much evidence to suggest the opposite, as Kathy Dopp originally suggested - and I also agreed, at the time. After some further analysis I found the evidence completely inconclusive. I would be astonished if there was not fraud in Florida in 2004, but there is no evidence that it divided along technological lines.
You welcome real-world evidence which proves that election was stolen, yet you fail to be impressed by the documented incidental data of switched and lost votes at the touch screens which heavily favored Bush.
Oh, I'm impressed. But it's not data you can do valid inferential statistics on.
You keep telling us that you have looked for fraud in the data and did not find it, and yet many other researchers have found the data to be highly circumstantial evidence of fraud.
And I have critiqued those analyses elsewhere, as has Mark Lindeman. I think theirs, like yours is flawed.
You tell us that exit polls are not true random samples and therefore one cannot use them for any probability analysis, yet the exit pollsters themselves claim a 1% margin of error with individuals randomly selected as they exited the polling station.
They did not. The "error" in the quoted margin of error refers to sampling error, which is not the only source of error in a survey. I have referred you many times to Mitofsky's own words about this, which as usual, you have chosen to ignore. So here they are again:
Mitofsky: I want to say a few words about reporting sampling error. A number of people who have spoken here have talked of not reporting sampling error because it was confusing all those dear mindless souls who listen to our results. They were concerned we would make people think that sampling error was the only error in the survey. I guess I am not too sympathetic with that point of view.
It appears that Mitofsky had reckoned without you.
You still believe that polls serve a purpose, but at the same time say that we cannot draw inferences from them - like the inference that Kerry won.
Yes, I think we can draw inferences from the poll data. I think it is enormously valuable data. But as with all data, inferences need to be drawn with care, and sources of error need to be modelled carefully. You have considered only one (or at most two) sources of error. There are many more.
You have not presented a comprehensive pre- and post-election state and national exit poll analysis. Nor have you analyzed the 2006 Generic polls. Nor have you analyzed the Bush job approvals. I have done analyzed them all. They are all in this thread.
This is macho strutting, TIA. "My analysis is bigger than your analysis". I have presented cogent analysis. One cogent analysis is worth any number of flawed analyses. Yours are flawed because you refuse to accept what anyone who knows anything about survey research knows, which is that survey data is vulnerable to both sampling error and non-sampling error. Ignoring the latter completely invalidates your analyses.
Where is your analysis? Instead of that scatter diagram, could you show us some actual NUMBERS?
The axes are marked. I have explained the units. The R squared is given. The N is 1250, as you could have inferred from the E-M report, but I did not make clear. You can compute the confidence interval of the regression line for yourself. You can also compute the correlation coefficient by taking the square root of the R squared value, and noting that the slope is slightly negative. You need no more in order to evaluate the analysis, and I have, in any case, interpreted it for you.
TIA, I suggest that instead you actually read my responses for once. If you have a substantive response, I shall, as always, be happy to hear it. I'm always happy to hear alternative ways of interpreting the data, and many hard-working DUers have helped me to figure out just what fraud might look like in the data, and how to frame testable hypotheses that might reveal it. But what you have posted not only completely evades the content of my posts, it mis-represents them. I shall put it down to ignorance this time. Next time I shall have to conclude you are lying.
edited for formatting and typo
|Printer Friendly | Permalink | Reply | Top|
|Home » Discuss » Topic Forums » Election Reform|
Software has been extensively modified by the DU administrators
Home | Discussion Forums | Journals | Store | Donate
Got a message for Democratic Underground? Click here to send us a message.
© 2001 - 2011 Democratic Underground, LLC