Well, it's finally happened. On Nov. 7 the long-running attempt by the media elite to take over the American electoral process was finally exposed. No one has yet suggested another War Between the States, but even a glance at the Gore-Bush electoral map of the United States should give you a shiver or two.

If the republic is to survive, a wooden stake must be driven through the heart of the whole misapplication of "scientific" statistical data analysis and sampling techniques by pollsters and the media elite.

In election after election, the media elite have declared one candidate a winner on the basis of one supposedly scientific exit poll. In election after election, they have claimed to be able to project on the basis of a few scientific exit polls of a few hundred voters, the votes of millions and millions of voters.

They're wrong. They've always been wrong. Florida has shown that they can't do it on the basis of one poll, in general, but if the election is close, they can't do it on the basis of any number of polls.

In order to apply statistical analysis to any "data" -- no matter whether obtained from the entire voter population or a small sub-group thereof -- the pollster has to identify a variable to which statistical analysis can be applied. For example, if the pollster asks a thousand randomly selected voters their age (or weight or height) their answer to that kind of question could constitute such a variable and statistical analysis can be applied.

Now, the pollster could ask a thousand randomly selected voters the question: "On a scale of zero to 10, how would you rate Al Gore as a potential president?" (That is the way many legitimate public opinion poll questions are structured.) "Potential presidential performance" would then become the variable and statistical analysis could then be applied to their answers. On the basis of that statistical analysis, the pollster could make a projection -- with some degree of confidence -- of the Gore performance-profile for the entire electorate.

However, a properly structured Gore performance-profile -- even if obtained as an "exit poll" -- doesn't even tell the pollster what percentage of his polled voters will vote for Gore, much less how the entire voter population will vote. After all, it is conceivable that more than a few voters, who think Al Gore would be an absolute zero as a president, voted for him anyway.

But if a pollster only asks a thousand randomly selected voters the question "Are you going to vote for Al Gore?" -- there are only three answers to that question: 1) yes, 2) no, 3) undecided. There is no way that a "mean" can be calculated for that "data" and, hence, no way to calculate a "deviation from the mean" and/or a margin of error.

By now you may be asking yourself if there is any way a pollster can sample the electorate so as to get some idea of how the entire electorate will vote? The answer is that there is a way to establish -- through polling -- a variable to which statistical sampling techniques can be applied.

Here's how: Ask a randomly selected group of voters if they are going to vote for Al Gore. Some percentage of the group will say that they are. Now you've got one data point. Ask another randomly selected group, and another, and another until you have got a hundred data points. Now you have a variable -- now you have a hundred values of that variable -- to which you can apply statistical analysis. Your variable is the percentage of each group that will vote for Gore. That hundred data-point Gore-profile will probably turn out to be bell-shaped, and now you're in business. You can start plugging your numbers into the formulas in the back of the statistical analysis handbook and you can project -- which some degree of confidence -- what percentage of the total electorate will vote for Al Gore.

But, applying statistical analysis techniques to the results of a single presidential preference poll of randomly selected voters -- as the news network pollsters have been doing -- is just dead wrong. There is no variable there. There are no data-values of a variable to statistically analyze. Therefore, a mean cannot be determined for the values of the variable and, thus, a standard deviation from the mean cannot be determined. As such, a margin of error cannot be calculated. So, no projections can be made as to how the entire electorate will vote.

Most pollsters know that, so they try to get around the necessary hundred or so random polls by attempting to select -- statistically -- a non-random "representative" group of the electorate and they poll that group once. But, even then, they are playing fast and loose with statistical analysis because the formulas in the back of their statistical analysis handbook apply to the selection of the representative group, not to what the voters in that group say they are going to do, or have done. In particular, there is absolutely no number that pollsters can derive from their one poll of their representative group of voters which can be legitimately plugged into the margin of error formula in the back of the book.

Up until now, there has been no way to catch the pollsters playing fast and loose with statistical analysis.

But then came Nov. 7 -- a date that will live in network election forecasting infamy.

What happened in Florida is that the pollsters convinced themselves and the media elite that they had developed a non-random representative sample of the Florida electorate. They decided, prematurely, on the basis of an "exit poll" of their "representative sample" of the Florida electorate, that Gore was going to easily win Florida. Their "degree of confidence" in their projection was so great that the media elite gave Florida to Gore before the polls in western Florida had even closed.

Now, even if the sample of the Florida electorate had been valid, it would not have been valid for them to project what that group did at the polls to the whole Florida electorate. The news networks should have known that it wasn't legitimate to do that, even then. But, the pollsters' representative sample turned out to be far from representative. Now, the news networks are claiming that they didn't know that it would have been illegitimate to make projections on the basis of one sample, no matter how representative it might have turned out to be. They plead ignorance. They say that they, themselves, didn't do anything wrong. They claim it was all the pollster's fault for not choosing a truly representative sample.

The media elite must not be allowed to get away with this. They must be forced to publicly acknowledge that the whole election result-projection process is fatally flawed and needs to be drastically overhauled -- or, better still, just plain junked.

The only valid way to continue this exit polling projection business is to apply statistical analysis techniques properly. That means taking hundreds of exit polls of randomly selected groups of voters. But suppose, as has been the case in recent years, that future elections are also close. Suppose that many final paper ballot tallies turn out -- as happened in Florida, New Mexico, Wisconsin, Oregon and Missouri -- to be something like 48 to 48 percent. Could the networks have "given" Florida to either candidate after the pollsters had taken even a hundred random exit polls? How about after taking a thousand samples?

We now know, in the specific case of Florida, the answer is no. The margin of error would be smaller after a thousand separate exit polls than after a hundred -- but even after a thousand exit poll samples of the 6 million ballots, the pollsters would have only looked at only a small fraction of the vote (and not looked at all at any of the absentee ballots). In Florida, several days after the election, the result is still somewhat in doubt even after all the 6 million votes have actually been counted and recounted!

Think about that for a moment. In a close election, even after properly applying statistical analysis techniques to a thousand groups (each composed of hundreds of voters), we now know the networks could not have picked a winner with any degree of confidence at all. The margin of error -- properly calculated -- would never have allowed it. But on Nov. 7, less than an hour before all the Florida polls had actually closed, all the TV networks "gave" Florida to Al Gore, all on the basis of a single exit poll.

Maybe the media elite won't have triggered another War Between the States after all, but someone is going to have a lot of explaining to do to Saint Peter for all that they put us through this past week.