(06-19-2019, 01:42 PM)flsprtsgod Wrote:

(06-19-2019, 12:31 PM)The Real Marty Wrote: Interesting, but I remember the question from college.  

How large a sample size is necessary for a population of 300 million, to get a confidence level of 95% that your margin of error is 3%?  

https://www.checkmarket.com/sample-size-calculator/

And the answer is, 1,068.   Use the sample size calculator above.  

In fact, 1,068 is a large enough sample size for any size population.  I remember the professor showing us this.  

Of course, the trick is to get a random sample, and that's where the real work comes in, but 1,068 is quite sufficient.

The math is based on random sampling. When you make a thousand calls to the Upper East Side of Manhattan you don't really get that.

That's why I said, "...the trick is to get a random sample, and that's where the real work comes in..."

No one could do 1,000 calls to the Upper East Side of Manhattan and call that a random sample without being charged with fraud. 

But if you can get a true random sample that reflects the entire population the survey applies to, 1,068 is plenty.

---