You are viewing an obsolete version of the DU website which is no longer supported by the Administrators. Visit The New DU.
Democratic Underground Latest Greatest Lobby Journals Search Options Help Login
Google

Reply #61: sigh.... [View All]

Printer-friendly format Printer-friendly format
Printer-friendly format Email this thread to a friend
Printer-friendly format Bookmark this thread
Home » Discuss » Topic Forums » Election Reform Donate to DU
Febble Donating Member (1000+ posts) Send PM | Profile | Ignore Sat Feb-10-07 05:21 PM
Response to Reply #59
61. sigh....
TIA
You have said YES to points 1-5. In fact, you have implicitly agreed to all points except #10.

6) Do you agree that we have now determined feasible vote share weighting "multipliers"?

Yes or No

FEBBLE
Well what do you mean by "feasible vote share weighting "multipliers" "? What are you multiplying? The only thing that makes sense is that you are "weighting" some cross-tabulation from the National Exit Poll, and you say below that "we have not said anything thus far about the National Exit Poll".

TIA
Yes, I am anticipating using the multipliers as weights for the NEP vote shares. And yes, I have not yet mention the National Exit Poll, so I will rephrase the question:

Do you agree that we have now determined feasible NEP vote share weighting "multipliers"?

Yes or No

I take your answer to be YES, since I was referring to the NEP, although I did not explicitly say so.


Thank you for the clarification. Well, the problem with your multipliers, is what you are multiplying with them. I certainly agree that there is an upper limit to the number of 2000 voters who can have also voted in 2004, and that this number is less than the total number of 2004 voters. That's as far as I go. I do not agree that we now have numbers with which we can validly multiply anything in the NEP data.

7) Do you agree that we have not said anything thus far about the National Exit Poll, much less sampling error or voter response?

Yes or No

FEBBLE
Well, what on earth are you multiplying ("weighting") if not the NEP?

TIA
Semantics. Of course, Im referring to the NEP. So lets move on.
Nothing to discuss here.


Well, apart from the fact that you just said you weren't discussing it. OK. Let's move on.

8) Do you agree that the ONLY remaining unknowns are Kerry and Bush vote shares of returning Gore, Bush, Nader and DNV2k voters?

Yes or No

FEBBLE
Well, assuming that what you are trying to calculate is the proportion of voters who voted for Bush or Kerry in 2004, no, of course I don't agree, because we have no way of knowing the vote shares for Kerry and Bush among those who didn't vote in 2000 either. That's a huge unknown, unless we are considering the exit poll data, which you just said we weren't considering at this point. On the other hand, if what you are calculating is the proportion of 2000 voters who voted for Kerry and Bush, then, yes, I agree it is an unknown - presumably that's why you are trying to calculate it.

TIA
Why the confusion? I calculate the ratio (proportion, share) of returning 2000 Gore, Bush and Nader voters to the 122.3mm total 2004 recorded vote. The remaining share must be the proportion of DNV2k voters to the total vote.

So I take your answer to be YES.


Sorry, I did misread this. Yes, I agree that 2004 voters can be partitioned into: voters who voted for Gore; those who voted for Bush in 2000; those who voted for Nader (or anyone else) in 2000; those who did not vote in 2000.



9) Do you agree that National Exit Poll vote shares can be considered to be the base case assumptions in calculating the National vote shares?

FEBBLE
Well, I suppose you have to start somewhere. I would have started with the NEP spreadsheet myself (and did), not the derived crosstabs.

TIA
I take your answer to be YES. NEP vote shares CAN be considered as the base case.

10) Do you agree that since there is a margin of error (which we need not argue about here) for the base case vote share assumptions, it makes sense to "stress test" the base case by analyzing alternative vote shares?

FEBBLE
I certainly don't consider that the total error in the NEP was limited to sampling error, and so I therefore would not agree that there was any inherent limit to any "stress test". In other words, demonstrating that only by "stressing" the data beyond the MoE could the data produce a Bush win would not demonstrate that a Bush win was "mathematically impossible" as you appear to conclude. It would merely demonstrate that if Bush in reality won, then the error in the poll was not limited to sampling error. For which there is abundant evidence, evidence you consistently ignore.

TIA
The Final NEP weights are mathematically impossible (not feasible). Both you and OTOH have already stipulated to that in the Game thread of August 2005. So any discussion of the How Voted weights is moot. We passed that hurdle a long time ago. We were in agreement then. So why are we still talking about false recall when it is no longer an issue. I thought we agreed on the the use of feasible weights a long time ago.


I don't know how to say this more clearly: THERE ARE NO "HOW VOTED" WEIGHTS IN THE NEP DATA. So of course discussion of them is moot - they don't ACTUALLY EXIST. THE NEP DATA WAS NOT WEIGHTED BY THE HOW VOTED QUESTION. It couldn't have been - the question was only presented to a quarter of the sample.

What I DID agree is that after weighting, the cross-tabulated proportions of Bush Gore voters could not possibly represent the actual Bush Gore proportions. This is blindingly obvious, and not in dispute. The reason that "false recall" is an issue is exactly as painstakingly explained by OTOH in post 52, and in Mark Lindeman's paper, to which I linked in post 55. But to save you one more mouse click, here it is again:

If respondents in the 2004 exit poll over-reported having voted for Bush in 2000 (and Mark Lindeman has assembled copious evidence that voters tend to mis-report having voted for the incumbent, four years later) then what you would expectto see in the 2004 crosstabs for 2000 vote is an overstatement of the the Bush vote. Which is what you DO see in the reweighted crosstabs. In other words, an alternative explanation to your fraud hypothesis is that people over-reported having voted for Bush in 2000, i.e. they tended to report having voted for Bush when they actually voted for Gore. Moreover, if the same real Gore voters who misreported having voted for Bush in 2000, ALSO voted for Bush in 2004 (and correctly reported having done so),this would not only result in an over-statement of Bush's vote in 2000, but it would also understate the Gore-Bush defection rate.

Really, TIA, this is actually simple math. You can do it if you try.


The original false recall argument was predicated on how the 2004 NEP respondents said they voted in 2000. Are you now saying that false recall also applies to how the respondents said they voted in 2004, just a few minutes after actually voting? Why would they lie about it?


No, see above. It is completely unnecessary to postulate that voters misreport their current vote. All that is required to reconcile the "Final" 2000 crosstabs with actual 2000 vote is to postulate that a small proportion of Gore 2000-Bush2004 voters misreported themselves as Bush2000-Bush2004 voters.

It should be obvious to anyone reading this thread that the false recall argument is a rotting carcass.


Well, it seems that nothing is obvious to you. But have one more go at actually reading the argument.

So now you must focus on the vote shares. In the Game thread, you and your buddy provided a vote share scenario forced to match the Bush recorded vote. But the votes shares are extremely implausible when put in juxtaposition to the Bush 48.5% rating on Election Day, the final 30-day undecided vote break to Kerry (60-38% based on the NEP), the many accounts of documented fraud in Ohio (including the recent recount convictions), the documented evidence of fraud in many other states. And to top it off, DNV2k and Nader 2000 voters were solidly for Kerry.


Well, as I've said, you are entitled to your own incredulity barriers. Frankly the best argument I can muster for massive fraud is an argument from incredulity. But that, as I've said, is not math, nor science neither. The evidence suggests that more Americans voted for Bush than didn't. And they seem to be regretting it now.

In light of all this, an impartial observer would clearly agree: any Bush win scenario is implausible and does not pass the smell test.


Smell tests are fine. Statistical tests are something else.

Know this: If Kerry won the popular vote in Ohio by 52-48%, as the documented vote-switching and spoiled vote evidence now indicates, he did better than 52-48 overall. The Ohio Democratic presidential vote share always trails the rest of the nation.


Well, I've never liked the smell of Ohio, but I don't see the evidence indicating 53-48% to Kerry. YMMV


TIA
11) Do you agree that the best way to "stress test" the base case is to employ a sensitivity analysis (two-way table) for the vote shares? In other words, to play what-if?

Yes, I agree that playing "what-if?" is the right approach. It's called hypthesis testing, and, in essence, it's what I do for a living. However, I do not agree that the plausibility limits to any such test can be computed from a calculation of sampling error. In the end, what limits the plausibility of your inference is your own credulity. Arguments from incredulity are not mathematical arguments.

TIA
I take your answer to be YES, playing what-if is the right approach.


Well, as long as you do it properly, of course, but yes, that's the essence of the scientific method.


FEBBLE
Your "search space" is bounded by plausibility assumptions that you do not state. There are many other "solutions" that you do not test because they would violate your implicit assumptions.

Any inference is only as good as your assumptions. There are implicit assumptions in your analysis that go far beyond the perfectly justified assumption "that the number of students who attended class on any given day was the simply the total who were registered in the class minus those who were sick or just decided not to show up."

You need to state what they are, and justify them. I do not consider them justified by the evidence.




TIA
And just what are those implicit plausibility assumptions?


I don't know. I'm asking you to state them. You gave some upthread.

What are the solutions I am not testing? Be specific. You are welcome to make your own assumptions as to the vote shares and turnout percentages.


The ones in Mark's paper, for a start.

I have provided you with a detailed solution space. Its the sensitivity analysis. Do you have another solution space? Then show us your scenario matrix. And tell us why yours is plausible.


Read Mark's paper.

The whole point of the sensitivity analysis is to provide a wide range of scenarios of alternative Gore and Bush voter turnout and Kerry and Bush vote share scenarios.


And yet you do not consider the possibility of misreported vote, and its implication for defection rates.


You have never shown us one sensitivity analysis. Why not? You just keep on displaying that same, lame scatter chart- over and over again. Is that the full extent of your analysis? You can do better than that.


There is absolutely nothing "lame" about that scatterplot. You haven't even begun to address the problem it raises for your analysis. You haven't even attempted to say why you consider it "lame". Similarly you have not attempted to explain why you fail to consider Mark Lindeman's scenario's. You simply scoff at them.

Enlighten us. Provide one plausible Bush winning scenario.


As I said, read Mark's paper. Then explain how your own scenario is consistent with my scatter plot.
Printer Friendly | Permalink | Reply | Top
 

Home » Discuss » Topic Forums » Election Reform Donate to DU

Powered by DCForum+ Version 1.1 Copyright 1997-2002 DCScripts.com
Software has been extensively modified by the DU administrators


Important Notices: By participating on this discussion board, visitors agree to abide by the rules outlined on our Rules page. Messages posted on the Democratic Underground Discussion Forums are the opinions of the individuals who post them, and do not necessarily represent the opinions of Democratic Underground, LLC.

Home  |  Discussion Forums  |  Journals |  Store  |  Donate

About DU  |  Contact Us  |  Privacy Policy

Got a message for Democratic Underground? Click here to send us a message.

© 2001 - 2011 Democratic Underground, LLC