By using this site, you agree to our Privacy Policy and our Terms of Use. Close
Epoch on 25 June 2009
Edit Thread
axumblade said:
Epoch said:
axumblade said:
Epoch said:
Lostplanet22 said:
They had a survey that counted of 700 people. They asked them for who you are gonna vote and Obama won; A survey of 700 people can be a very good indication for whole the country.

WOW.  You just boggled my mind. 

I once flipped a coin 701 times and my "survey" told me that Bush would win in 2000.  Guess my coin was a pretty good indicator for the whole country as well -.-

It's called a survey for a reason...And it was slightly accurate, your argument makes no sense considering it was a sample survey and in all likelhiood, they interviewed people from different states for their answers. Just because not everybody liked Obama doesn't mean that he wasn't going to win. It's about who has the most votes, not who has all of the votes...

You both missed the point of my example.  My example was simply to demonstrate that the likelyhood of their "survey" being correct was about 50%. 

700 people surveyed is hardly indicative of anything.  The fact that the survey was right is due more random chance than anything else. 

In short: small data sets used to represent much larger populations are often rife with problems. There are countless examples of this.  Look some up

Yeah...the thing about that is that nobody is going to give the whole populous a survey unless it's conducted by someone who has a lot of time/money on these tests. I don't think I understand what you're saying because you're pushing a personal belief on a sample survey. I suppose it could work as flipping a coin would though based on the fact that Obama and McCain were the only ones really left in the competition at the end, but I highly doubt that's what you're even trying to say.

Actually thats exactly the point I was trying to show.  If there are 2 equally athletic people about to engage in a footrace, and you ask 700 people who is going to win, the survey will turn out to be right approximately 50% of the time. Its hardly mind blowing stuff, and shouldn't be used to vindicate small data sets of being generally poor at representing whole populations.

My "personal belief" that small data sets are often poor representations of much larger populations is widely held by most statisticians as well.  There are things you can do to mitigate the problems inherent in small data sets, but you should always check your results with known quantitifiable data to avoid looking foolish. 

Anyhoo, my point is simply that you should take any survey, or poll's, results with a grain of salt.  Particularly when we can extrapolate their results into our "known" sales data, and find it off by large margins.

If you are actually interested in statistics, perhaps a course on the subject would help your understanding.

But this discussion is now way off topic.