Lawlight said: 2000 data points for worldwide numbers? Well, that would expect that big inaccuracies. |
If we assume a reasonably unbiased sampling for those 2000 data points, then you would expect a margin of error at 95% confidence of 2.2%, assuming an infinite population. Smaller for the real population. That is, 95% of the time you would expect the random sample's mean to be within 2.2% of the real mean.
Seriously, if you don't understand basic statistics, don't get involved in this discussion, you're in over your head.