And that's the point. People are still talking about that Apple ad 3 decades later. How many of today's ads will be remembered in 30 days, much less 30 years?
I would suggest that you get a book on statistics and learn something before embarrassing yourself further.
Your statement applies ONLY IF the sample is random and representative. If a sample is random and representative, then increasing the sample size increases the probability that the observed result is the same as the population as a whole. The numbers are controlled purely by mathematics and can easily be looked up. Just from memory, but the order of magnitude will be right: If you have a sample size of 1,000, there will be a roughly 95% probability that the result will be within +/- 3% of the result for the entire population.
That does NOT apply if the sample is not representative (as, for example, in this case where you have a self-selected sample). Example: let's say that I want to know what percentage of the world's population speaks Mandarin or Cantonese. If I sample 1,000 people in Beijing, the result will be close to 100%. If I sample 1,000 people in Little Rock, Arkansas, the result will be a few percent, at most. The problem is that neither sample is representative of the world's population.
As I've pointed out repeatedly in the survey numbers presented here on various topics, the numbers are often worthless, not because of sample size (since, properly done, a sample of even a thousand can give you useful results) but because samples are not representative of the population as a whole.
I'll make the same suggestion to you although you can't save yourself from looking like an ass. The damage is already done. What I said was absolutely correct. With a sample size above 30 the chance that the sample mean differs from the population mean rapidly approaches zero. As sample size increases it is increasingly unlikely to pick a sample whose mean is some number of standard deviations away from the population mean (it's a bit more complicated than that but for the purpose of this argument it's all you need to know). That statement accounts for the fact that you could potentially choose a sample that does not reflect the population. What my statement also says is that your chances of selecting such a population are exceedingly low, especially when your sample is several thousand people. The chances of a skewed sample of that size is virtually zero.
I also want to emphasize the fact that the population is all viewers of the video, not all people in the US, all people in the world, all smartphone users, or any other group.
Edited by wakefinance - 2/2/13 at 11:23am