To the Editor:
In your Nov. 4 article regarding a Presidential election poll taken over the weekend, you stated that "64 percent of [Dartmouth] students support President Clinton." I dispute this statement on a statistical basis.
First of all, the poll was conducted over electronic mail and required an active response. This removes the randomness from the sample, and without a random sample it is statistically impossible to make any conclusions about the population at large. Maybe the 1800 people selected for the poll were chosen randomly, but the 532 people who responded were certainly not a random subset of the 1800. Some people surveyed may have been away for the weekend; others may make a practice of refusing to respond to impersonal blitzes. Not every surveyed student was equally likely to respond, and thus the sample is not random and tells nothing significant about the population.
In addition, no acknowledgment of the margin of error was made anywhere in the story, with regard to this or any poll. A statement of standard error is obligatory when making inferences about an entire population from a random representative sample, and is even more important in interpreting the results of a flawed experiment like this one. Without knowing how accurate the sample results are, we have NO MATHEMATICAL WAY of knowing what the response of the whole population will be. Thus it is inaccurate, and statistically irresponsible, to make any statements about the opinion of the student body at large based on the results of this poll.
