Originally posted by DeDolfan
Granted, but what about these other so called scientific polls like Gallup and such? I think that polls can be "made" to say whatever you want. "Random" samplings can be easily fixed, ie if you want to show favorable #s on conservative issues, take your poll outside the GOP convention!
But, that's a flawed sample and wouldn't be scientific. At least two reasons: (1) "random", within the context of statistical sampling does not mean, standing outside some place and asking whoever you happen to see; (2) even if it were conducted in a scientific sampling, if the context of the response is not placed in terms of "x% of republicans state that they would y", then it's a deceptive poll, and not scientific because of the bias.
[/B][/QUOTE] phone polls are another that crack me up. Obviously they have to do them but I have never or ever heard of anyone personally being polled by phone, other than a local thing done during election times. Polls are made to sway the fence sitters anyway. :goof: [/B][/QUOTE]
As far as "sampling" goes, a survey taken by telephone may be more likely to provide an accurate result than the example you just provided conccerning GOP convention. However, you can't even just say: "I need 1400 people, there are 140,000 listings in the telephone book, so I'll just take ever 100th name, call them and ask them. There are flaws that will skew even that approach to sampling - it's not truly random. To be random, each person in the population that is being studied must have an exactly equal chance of being selected as anyone else. Thus, an internet "poll" is going to get multiple votes from people with an ax to grind, is only going to sample people who feel motivated enough to "vote", has absolutely NO way of determining what the population is that it is sampling (is it Dolphin fans? Any NFL fans? Anyone on the internet? No way of knowing). Scientific surveys with carefully crafted questions can be accurate and with a relatively small number of participants, can provide be extrapolated to the focus population (the group of people that you want to be able to extrapolate to - not everyone in general).
The goal of a true scientific survey is to eliminate bias in the sampling and in the question. If it's trying to push an agenda, then it's not "science", but manipulation.
With the Sun-sentinel question - it asks "Do you think Brian Griese should start Monday night?". Even if the sampling was legitimate, even if there was a definied population, the question itself does not tell you whether the people who vote yes are voting so because they believe Fiedler is too injured to play, because Fiedler could play but his lack of mobility would hurt his game, because they think BG is a better choice for all games, regardless of "Monday night" .... see? There are dozens of flaws in these internet polls, probably many more than that even, but do you see why these things can't even be generalized to represent the opinion of 1700 people who voted? (there's not even any control over someone voting multiple times, so you can't even say that the results reflect what 1700 people think, much less generalize it out to Dolphin fans, NFL fans, k-mart shoppers or whatever undefined population is being sampled.