Gnizmo on 02 September 2009
| Kantor said: A 12,000 sample size might be accurate for, say, 5 million products. For 100 million I'd rather go with manufacturer figures. This is marginally more accurate than the GamePro poll. The ratio of sample size:install base is roughly 1:8000. This is equivalent to an install base of 80,000, and asking TEN people. |
Statustics coesn't work that way. Sample size does not need to scale with population size. All you have to do is have a sufficient large number such that you are almost guaranteed (95-99% chance) to get a "representative sample." Supposing there is no selection bias 12,000 would be more than enough to accuratley give a representative sample from a population of 100 million. You have absolutely no idea what you are talking about. Please go read a statistics book.
